DJ ASCII -18

DJ ASCII-18

DJ ASCII-18

Editor-in-chief Dr. Hari Vasudevan

Convener DJ-ASCII

Editor Dr. Narendra Shekokar

Joint-Convener DJ-ASCII

Second Impression: June 2018

© D.J. Sanghvi College of Engineering, Mumbai, Maharashtra

DJ ASCII-18 ISBN: (Under Process)

No part of this publication may be reproduced or transmitted in any form by any means, electronic or

mechanical, including photocopy, recording, or any information storage and retrieval system, without

permission in writing from the copyright owners.

DISCLAIMER

The authors are solely responsible for the contents of the papers compiled in this volume. The publishers

or editors do not take any responsibility for the same in any manner. Errors, if any, are purely unintentional

and readers are requested to communicate such errors to the editors or publishers to avoid discrepancies in

future.

Published by

Dept. of Computer Engineering & Dept. of Information Technology, SVKM’s Dwarkadas J. Sanghvi College of Engineering, VileParle (West), Mumbai: 400056. Website: www.djsce.ac.in

Preface

We are pleased to present the proceedings of the State Level Project Competition DJ ASCII. The

competition was jointly organized by the departments of Computer Engineering and Information

Technology of SVKM’s Dwarkadas J. Sanghvi College of Engineering on 13th April, 2018 in

Mumbai, India.

The main aim of DJ ASCII was to provide a platform for budding engineers and researchers from

all over Maharashtra to share and demonstrate their innovative ideas in the field of latest

technology. We are very happy to say that DJ ASCII has achieved what it had aimed for by

receiving an overwhelming response of total 150 participations, out of which 95 were selected

after a rigorous review by our panel of expert reviewers. 74 project groups registered for the

competition.

The competition received projects primarily in the domains, which were not limited to, Artificial

Intelligence, Computing, Human Computer Interactions, Data Mining and Analytics and Network

& Security. The projects were presented and demonstrated by students from various engineering

institutes in parallel sessions. Students also had presented their project ideas with technical paper

and abstract of those papers are published with ISSN number.

We thank our patrons Shri. Amrish R. Patel (President, SVKM), Shri. Bhupesh R. Patel (Joint

President, SVKM), Shri. Bharat M. Sanghvi (Vice President and Trustee, SVKM, In-charge,

DJSCE), Shri. Sunandan R. Divatia (Hon. Secretary, SVKM), Shri. Jayant P. Gandhi (Hon. Joint

Secretary, SVKM), Shri. Shalin S. Divatia (Hon. Joint Secretary, SVKM), Shri. Utpal H. Bhayani

(Hon. Treasurer, SVKM), Shri. Harshad H. Shah (Hon. Joint Treasurer, SVKM) and Shri. Harit

H. Chitalia (Hon. Joint Treasurer, SVKM) for their valuable guidance and support. We are

extremely grateful to our Management SVKM for their wholehearted support in organizing DJ

ASCII.

We thank the advisory committee for their guidance and inputs throughout the organization of the

competition. The review committee was very helpful in providing timely and constructive reviews

of the papers. We would also like to thank all the students, who participated and showed active

interest throughout the competition.

Finally, we thank our fellow members of Technical and Organizing Committee and our student

volunteers for the smooth conduct of the competition. Their sincere efforts and contribution have

certainly made a huge impact on the success of this event. We hope you will have a wonderful

experience going through the proceedings.

Thank you very much.

Warm regards and good wishes!

Dr. Hari Vasudevan, [Convener, DJ ASCII]

Dr. N.M. Shekokar, [Joint-Convener, DJ ASCII ]

Committee

Convener Dr. Hari Vasudevan

Principal

Joint-Convener Dr. N.M. Shekokar

Head, Department of Computer Engineering

Advisory Committee Dr. A.C. Daptardar Vice-Principal (Admin.)

Dr. Manali J. Godse Vice Principal (ACAD) and Head, Department of Biomedical Engineering

Dr. V. Ramesh Professor and Head, Department of Chemical Engineering

Dr. A.A. Deshmukh Professor and Head, Department of Electronics and

Telecommunication Engineering

Dr. K.N. Vijay Kumar Professor and Head, Department of Mechanical Engineering

Dr. P.S. Joshi Head, Department of Electronics

Prof. R.S. Khavekar Training and Placement Officer

Technical and Organizing Committee

Dr. Abhijit R. Joshi Dr. Meera Narvekar Dr. Neepa K. Shah

Dr. Ram Mangrulkar Prof. Vinaya N. Sawant Prof. Aruna U. Gawade

Prof. Kiran Bhowmick Prof. Kriti Srivastava Prof. Purva P. Raut

Prof. Khushali P. Deulkar Prof. Neha A. Katre Prof. Lakshmi D. Kurup

Prof. Harshal D. Dalvi Prof. Ashok P. Patade Prof. Arjun K. Jaiswal

Prof. Chetashri S. Bhadane Prof. Ruhina B. Karani Prof. Anusha Vegesna

Prof. Sindhu S. Nair Prof. Stevina Correia Prof. Mitchell R. D’silva

Prof. Lynette R. D’mello Prof. Pranit Bari Prof. Deepika Dongre

Prof. Sudhir Bagul Prof. Pankaj Sonawane Prof. Pratik Kanani

Prof. Dinesh Tharwani Prof. Suchita Rane Prof. Amruta Patil

Prof. Priya Lande Prof. Nancy Nadar Prof. Sonali Jadhav

Prof. Archana Nanade Prof. Amit Sahu Prof. Ankita Kadu

Project Reviewers’ Committee

Dr. Amiyakumar Tripathi

Dr. Sunil Surve

Mr. Sachin Kadam

Mr. Pinaki Bhowmik

DJ ASCII – 18

_____________________________________________________________________________________

i

Contents

1. Detecting Objectionable Content in Video Clips for Children .................................................. 1

Rushabh Dharia, Chirag Jain, Niki Jain

2. Real Time Face Recognition for Banking Security System ........................................................ 1

V.K.N Kamlesh Pai, Manoj Balraj, Sachinkumar Mogaveera, Deepak Aeloor

3. A Crop Suitability Predictor Using Supervised Learning ......................................................... 2

Zeel Doshi, Subhash Nadkarni, Rashi Agrawal, Prof. Neepa Shah

4. Accident Detection, Prevention and Reporting using Raspberry Pi ......................................... 2

Vighnesh Desai, Suraj Hazraj, Pavan Hebli, Nikhil Kore

5. Profanity Free Facebook ............................................................................................................... 3

Ameya Kasbekar, Rashmi Rana, Vidhi Shah, Dr.Abhijit R. Joshi

6. Detection of Woman Harassment in a Train Compartment ...................................................... 4

Murtaza Rajagara, Dhwani Shah, Ansh Thakkar

7. A Self-Driving Model for Low Cost Implementation ................................................................. 4

Yash Trivedi, Prashil Negandhi

8. Movie Genre Classification using Poster ..................................................................................... 5

Akshat Barbhaya, Karan Gada, Mahadevan Kounder

9. Automated Fake News Detection .................................................................................................. 5

Shreya More, Hrishikesh Telang, Yatri Modi

10. Artistic Style Transfer Using Deep Learning .............................................................................. 6

Pankil Daru, Siddhant Gada, Meet Chheda, Prof. Purva Raut

11. An Offline Handwriting Recognition Approach for Ruled Pages using CNN and LSTM ...... 7

Jay ashok lal, Meet mukadam, Viral pasad

12. Nutritional Estimation from Fast Food Images and Alternative Recipe Suggestion for

Diabetic Patients ............................................................................................................................ 7

Kaiz Merchant, Yash Pande, Ashutosh Singh, Prof.Mr. Pankaj Sonawane

13. Smart Health Monitoring .............................................................................................................. 8

Shweta Pujari, Sneha Salunke, Darshana Samala, Nilofer Shaikhr

14. Autonomous Car ............................................................................................................................ 8

Onkar Walavalkar, Suraj Parab, Vikas Tiwari

15. Citizen Service Problem (CSP) ..................................................................................................... 9

Aditya Kapadia, Lucky Rathod, Jigar Panchal

DJ ASCII – 18

_____________________________________________________________________________________

ii

16. Automated Essay Generation ....................................................................................................... 9

Dhruvil Mehta, Rajat Bhagat, Aditya Dubey

17. Real Time Surveillance System on Hadoop Image Processing Interface ................................ 10

Prof. Er. Mohammed Ahmed Abdul Mannan, Shamuwel Ahmed Ansari, Shebaz Ansari, Sayed

Adnan Husain

18. Voyageur- A Smart Trip Planner ............................................................................................... 10

Akshen Kadakia, Urvi Mistry, Devanshi Desai, Prof. Mitchell D’silva

19. Smart Meter Data Compression and Pattern Extraction ........................................................ 11

Ankit Shah, Shweta Sunderkrishnan, Dhrumil Dhutia, Prof. Aruna Gawade

20. Intelligent Alzheimer’s Detector using Deep Learning ............................................................. 12

Mukul Puranik, Himanshu Shah, Keval Shah

21. Design and Implementation of Neuromuscular ........................................................................ 13

Amit Katariya, Shantanu Madiwale, Meet Mehta

22. Human Understanding Analyzer (Machine Learning) ............................................................ 13

Manogya Prasad, Rizwan Japanwala, Harshil Vora, Prof. Mrs. Lakshmi Kurup

23. Heart Rate Evaluation and Risk calculation ............................................................................. 14

Chintan Devda, Jaydeep Gami, Meet K Mehta

24. Text to Image Generation using Deep Learning ....................................................................... 14

Sudhir Bagul, Kumpal Dhruv, Nikhil Kamat, Pooja Kulkarni

25. Calamity Evacuation using Social Network .............................................................................. 15

Parth Vora, Mili Desai, Aditya Vallat

26. IoT Based Intelligent Medical System-IMED ........................................................................... 16

Yashraj Kotian, Pawankumar Dubey, Romil Badami,Vivek Giri, Dr. Rekha Sharma

27. On Panel Auxiliary Warning System for Indian Railways ...................................................... 16

Saurabh S. Deone, Harshal D. Waje, Pravin Kanase

28. Real Time Drowsiness Detection System ................................................................................... 17

Mr. Shailesh Sangle, Bharat Rathore, Rishabh Rathod, Aakashkumar Yadav

29. Triple Technique Diagnosis Using Machine Learned Classifiers ............................................ 18

Het Sheth, Dhruman Shah, Mihir Gada, Purva Raut

30. On Road Activity Monitoring System ........................................................................................ 18

Tirth Patel, Viral Shah, Vipul Hodge, Raj Desai

31. Solar Energy prediction using Artificial Neural Network ....................................................... 19

Pranali R. Mane, Vivek H. Patel, Saurav S. Yadav, Dr.Abhijit R.Joshi

32. Audio Recording and Analysis to Detect and React to Anthropogenic Disasters .................. 20

DJ ASCII – 18

_____________________________________________________________________________________

iii

Shalin Parikh, Raunak Vijan, Zaid Merchant, Prof. Mr. Pratik Kanani

33. Identification of Potential Cyberbullying Tweets using Hybrid Approach in Sentiment

Analysis ......................................................................................................................................... 21

Akankshi Mody, Reeya Pimple , Shreni Shah, Dr. (Prof.) Narendra Shekokar

34. Application of ML Techniques for the Analysis of Hypertension and Prediction of Vein

Function in Hemodialysis ............................................................................................................ 21

Dr. Gresha Bhatia, Mihir Wagle, Neeraj Jethnani, Juhi Bhagtani, Aishwarya Chandak

35. Prediction of Personality based on Handwriting ...................................................................... 22

Sharanya Ojha, Abhishek Shah, Sagar Shrinagarpure

36. Multiclass Classification of Imbalanced DataStream ............................................................... 22

Pratik Parekh, Medha Shah , Utsav Shah

37. Dermatological Disease Classification ........................................................................................ 23

Sagar Dedhia, Pratik Jayarao , Yash Doshi

38. Machine Learning Approach to Foretell the Probability of a Crop Contracting a Disease . 23

Viraj Mehta, Chahat Jain, Karan Kanchan

39. Medical Intelligent Record Assistant ......................................................................................... 24

Suhail Barot, Dhanashree Jatar

40. Automated Evaluation of Subject Answers using Text Processing and Sentence Similarity 25

Aditya Maniar, Devansh Gada, Chantelle D’Silva

41. Game Automation Using Reinforcement Learning .................................................................. 25

Bhargav Mehta, Rohan Madhani, Viral Lakhani

42. Autonomous Driving Car using Neural Networks .................................................................... 26

Raj Chaudhari, Shivani Dubey, Jayesh Kathale

43. Image Caption Generation Using Deep Neural Networks ....................................................... 26

Siddhant Agarwal, Himani Deulkar

44. Simulation of Self-Driving Car in a Simulated Environment .................................................. 27

Atharva Kunte, Avasyu Bhatia, Ziyad Dhuka

45. Safe Driving System ..................................................................................................................... 27

Abhishek Shukla, Yeshwant Ranka, Bhavya Shah, Prof.Mrs. Sindhu Nair

46. NaYaNa – A Phonetic script to Make Communication Easy ................................................... 28

Prachi Rahurkar, Deepa Ramrakhiani, Smriti Rao

47. Context Based Question Answering System .............................................................................. 29

Avais Pagarkar, Rudresh Panchal, Swapneel Mehta, Lakshmi Kurup

48. ALLIO-All in One Multithreaded Server .................................................................................. 29

DJ ASCII – 18

_____________________________________________________________________________________

iv

Darsh Shah, Vaibhav Shah, Sumit Busa

49. Security and Privacy for Data on Cloud using Image Sequencing and Multiple Encryption

....................................................................................................................................................... 30

Zambare Jayesh Chhabildas, Bedekar Ganesh Ramchandra

50. Lifestyle based Disease Prediction System................................................................................. 30

Harsh Jain, Jay Jain, Praful Kothari, Prof. (Mrs.) Chetashri Bhadane4

51. Data Preprocessing for Efficient Sentimental Analysis ............................................................ 31

Shreyas Wankhede, Ranjit Patil, Sagar Sonawane

52. Personalized Travel Sequence Recommendation on Multi-Source Big Social Media ........... 31

Tarun Reddy, Jenila Sanghvi, Deval Vora

53. Anomaly Detection in Legal Documents .................................................................................... 32

Partik Aher, Kunal Doshi, Tushar Dey, Prof. (Mrs.) Purva Raut

54. Understanding Short Texts through Semantic Enrichment and Hashing .............................. 33

Devashree Patil, Nitesh Jadhav, Kanan Muthe

55. FOREX ......................................................................................................................................... 33

Nimit Parekh, Manav Shah , Varun Shah

56. Document Clustering Using Ontology........................................................................................ 34

Maithili Shah, Parth Mehta

57. Smart Health Prediction .............................................................................................................. 34

Malav Shah, Harsh Shah, Parth Oza, Dinesh Tharwani

58. Twitter Analysis of Trending Hashtags ..................................................................................... 35

Utkarsh Dubey, Varun Agarwal, Bhavya Shah

59. Classifying Imbalanced Data Streams in the Presence of Concept Drift ................................ 36

Chandrasekhar Raman, Pratik Bhambhani, Aqsa Bhimdiwala

60. 3D Visualization and Colorization of Brain MR Images .......................................................... 36

Dhruvesh Mehta, Gitika Daswani

61. Attention Monitoring System in a Classroom ........................................................................... 37

Uma Sreeram, Aastha Joshij, Nikita Parmar

62. Lung Cancer Detection ................................................................................................................ 38

Dhyanvi Jhaveri, Manali Nagda

63. Oculus Vision for Blind ............................................................................................................... 38

Dennis Jayesh Mistry, Varun Vasudev Mukherjee, Nayan Thakor Panchal, Prof. Aruna U. Gawde

64. Optic Interactive System ............................................................................................................. 39

Bhuvana Iyer, Krishnakant Thakur, Sanjana Kale, Shivani Jadhav

DJ ASCII – 18

_____________________________________________________________________________________

v

65. Smart Trainer: A Virtual Guide for Personal Training in Exercise ....................................... 39

Twinkle Dhanak, Saheb Singh, Niyati Maheshwari

66. VR HARAPPA ............................................................................................................................. 40

Hem Acharya, Prutha Dubhashi, Rhea Sanghvi, Prof. Neha Katre

67. An Educational Augmented Reality App to Enhance Learning Experience .......................... 40

Vineet Malhotra, Dhruvil Desai, Hasan Banswarawala

68. VR Chem Lab ............................................................................................................................... 41

Chintan Shah, Sharang Ukidve, Neal Gosalia, Prof. Neha Katre

69. Malware Generation using Taint Graph ................................................................................... 41

Ayushya Mishra, Mehul Sanghavi, Pawan Shah, Aman Gandhi

70. AWI Automatic Website Developer ........................................................................................... 42

Mahavir Rathod, Aayushi Shah, Karan Vyas, Mitchell D’silva

71. Load Balancing in Software Defined Network .......................................................................... 42

Ashwini Swain, Karan Savla, Kushal Ajmera

72. Data Encryption using Fibonacci Series and Unicode Characters .......................................... 43

Prof. Prasad Tambekar 1, Prof. Vijay Shelke 2, Snigdha Mandal 3, Snehal Khachane4, Diksha

Shetty 5

73. No Cost Real-time Content Management System ..................................................................... 43

Emmanuel Francis Kolengaden

74. Lightweight Authentication and Encryption Mechanism in Routing Protocol for Low Power

and Lossy Networks ..................................................................................................................... 44

Akshay Shah, Nishit Sakariya, Dishith Poojary

DJ ASCII – 18

1

PAPER ID: AI02

Detecting Objectionable Content in Video Clips for Children Rushabh Dharia1, Chirag Jain2, Niki Jain3

1,2,3, Department of Computer Engineering,

Dwarkadas J. Sanghvi College of Engineering, Mumbai, India

E-mail: [email protected], [email protected], [email protected]

This paper describes our project “Detecting Objectionable Content in Video Clips for Children” which uses Deep Learning to find out if a video file is suitable for children to watch. Exposure of objectionable content like nudity, violence and drug abuse can affect children. Our software can detect such content beforehand by analyzing the frames in the video. Deep learning has been demonstrated to achieve excellent results for image classification and object detection. However, the impact of deep learning on video analysis has been limited due to complexity of video data and lack of annotations. In this paper, we propose a recurrent convolutional network called RCNN for action detection in videos. The proposed architecture is a unified deep network that is able to recognize and localize action based on 3D convolution features.

PAPER ID: AI03

Real Time Face Recognition for Banking Security System V.K.N Kamlesh Pai 1, Manoj Balraj2, Sachinkumar Mogaveera3, Deepak Aeloor4

1,2,3,4 , Department of Computer Engineering,

St. John College Of Engineering and Management Village Vevoor, Manor Road, Palghar (East) 401404

E-mail: [email protected], [email protected], [email protected], [email protected]

Face recognition is a common biometric authentication technique used to analyse the face images and extract useful recognition information from them, which are always called as a feature vector that is used to distinguish the biological features. Face Recognition process begins with extracting the coordinates of features such as width of mouth, width of eyes, pupil, and compared it with a stored face template. The aim of the proposed system is to design an autonomous security system that performs face recognition based surveillance combined with a hardware mechanism to lockup the secured region. Haarcascade algorithm is used to detect and extract the face from an image thereby storing samples in order to train the system. This system consists of a buzzer alarm and two cameras diagonally placed. The camera locates, tracks people entering the secured room, recognize the individual and message is passed to the control room which is stored in the log file. Any unauthorized access is logged along with a buzzer alarm to notify the control room followed by locking the exit points of the system. This system focuses on system security using face recognition which can be installed at banking suits.

DJ ASCII – 18

2

PAPER ID: AI04

A Crop Suitability Predictor Using Supervised Learning Zeel Doshi1, Subhash Nadkarni 2, Rashi Agrawal 3, Prof. Neepa Shah4

1,2,3,4, Department of Information Technology,


E-mail: 1 [email protected], 2 [email protected], 3 [email protected] 4 [email protected]

A vast majority of the Indian population depends either directly or indirectly on agriculture for their livelihood and survival. Consequently, it is ineluctable that agriculture plays an imperative role in the country. The general rule of thumb followed by most of the farmers is to depend on their intuition to decide which crop to grow in a particular sow-season. This approach, undoubtedly, is flawed, as a single farmer cannot be expected to analyze all the contributing factors of crop growth before determining which one to grow. A misguided or imprudent decision by the farmer can have negative repercussions on both himself as well as the agricultural economy of the region. To help alleviate this issue, we present an intelligent model that would serve as assistants to the Indian farmers. It would take into account the geographical location of the farmers during a particular season to help him determine which crop would be the most optimal for growth. This prediction would be based on environmental factors such as temperature, rainfall and soil pH and nutrient content, all of which play a significant role in determining crop yield.

PAPER ID: AI05

Accident Detection, Prevention and Reporting using

Raspberry Pi Vighnesh Desai1, Suraj Hazraj2, Pavan Hebli3, Nikhil Kore4

1,2,3,4 , Department of Electronics and Telecommunication,

SIES Graduate School of Technology Nerul, Navi Mumbai.

E-mail: [email protected]

The main concern of Public Safety Organization (PSO) around the world is to minimize the effect of vehicle accidents, aiding as many injured people as possible and providing 24/7 assistance on the accident spot. But in India, the problems that many rescue teams face is to reach the accident spot on time and helping the injured. The emerging technologies in the mobiles phones and Internet of Things can help to save a lot of lives. Hence a smart, reliable IoT system for accident detection, prevention and reporting is designed using Raspberry Pi. The proposed system not only can detect accidents but also can prevent them beforehand. This is a novel system which can aid the rescue teams and thus prevent accidents by saving lives.

DJ ASCII – 18

3

PAPER ID: AI06

Profanity Free Facebook Ameya Kasbekar1, Rashmi Rana2, Vidhi Shah3, Abhijit R. Joshi4

1,2,3,4,, Department of Information Technology,



According to June 2017 statistics there are over 2.01 billion monthly active Facebook users worldwide. Facebook as a social media website provides a platform for its users to share photographs, events, current news and trends with their friends and family located worldwide as well stay updated about the ongoing events in other people’s life. This platform allows them to freely express their thoughts about the ongoing scenarios in other people’s life. Hence it has proven to be a boon as well as a misfortune at the same time. Typical age of a Facebook User ranges from adolescents to adults. As per "Global Threat Report", 65% of content on Facebook is offensive. Facebook currently has content reviewers who spend only a few seconds to judge each post. Mark Zuckerberg, the CEO of Facebook, recently stated “No matter how many people we have on the team, we’ll never be able to look at everything”. Hence we aim to develop a project to automate the task of reviewing written posts and comments, and further categorize them as offensive or non-offensive. First step in categorization identifies the presence of any offensive word in the comment or phrase whereas the second step involves classifying the comment by analysing the relation between the words in the sentence. Finally if the results of categorization are positive, i.e. offensive then the corresponding content will be hidden, thus restricting users from reading it. In this way by implementing this project we are trying to overcome Cyberbullying.

DJ ASCII – 18

4

PAPER ID: AI11

Detection of Woman Harassment in a Train Compartment Murtaza Rajagara 1, Dhwani Shah2, Ansh Thakkar3


K. J. Somaiya College of Engineering, Mumbai, India.


Women harassment is an issue that has spread far and wide. The number of reported cases are on the rise in the country and all over the world which is a growing concern for our society. The purpose of our system is to detect harassment which is captured through the surveillance camera and alert the concerned authorities so they can take the necessary actions. The model of the proposed system consists of two folds, the first part uses dense optical flow algorithm to verify any abnormal activity. The second part of system consists of pose detection which is then classified using action recognition algorithm. On detection of harassment, the system triggers an alarm to alert the authorities.

PAPER ID: AI12

A Self-Driving Model for Low Cost Implementation Yash Trivedi1, Prashil Negandhi2

1,2, Department of Computer Engineering,



The key motivating factor for the underlying research work is to tackle the autonomous driving problem. The proposed research work attempts to design a model to effectively bridge the gap between manual and autonomous cars. It uses computer vision and machine learning to identify obstacles like other vehicles and pedestrians, and respond correctly to traffic-signals and traffic-signs. For training the self-driving model, different types of neural networks will be tested. The authors will attempt to find if a low cost autonomous driving vehicle can be developed using this approach. This would be very advantageous to quickly and conveniently introduce autonomous vehicles in any area.

DJ ASCII – 18

5

PAPER ID: AI13

Movie Genre Classification using Poster Akshat Barbhaya1, Karan Gada2, Mahadevan Kounder3

1,2,3 Department of Computer Engineering,


Judging a book by its cover is an old adage that warns against evaluating the merit of something based strictly on its outward appearances. However, taken literally, we set out to see if we can in fact judge a book by its cover, more specifically judge a movie by its poster. This paper presents a method for movie genre categorization of movie poster. Our approach decomposes each poster into feature and its descriptor and cluster this descriptor into Bag-of-Visual-Words (BOVW). We approach the genre classification task by mapping BOVW into movie genres like action, romance, comedy, drama or horror movies using learning algorithm. On one hand we will explore various algorithms like SIFT, SURF etc. for feature extraction and cluster them using K-Means algorithm to obtain BOVW from the object recognition of poster of the movie and on the other hand we will test various learning algorithm like Neural Network(NN), SVM etc. to obtain movie genre as output from this BOVW.

PAPER ID: AI14

Automated Fake News Detection Shreya More 1, Hrishikesh Telang2, Yatri Modi3

1,2,3, Dwarkadas J. Sanghvi College of Engineering, Mumbai, India


The discrimination between truth and falsehood has received significant attention from various fields such as philosophy, psychology and sociology. Fake news has been at the centre of the debate raging about the outcome of the 2016 US Presidential Election and the rise of social media has made it easier to blur the line between what is and what isnt a reliable source of information. Millions of people use social media frequently, making the spread of fake news a pressing issue. Recent advances in NLP made us think about approaching this issue from a data-driven perspective. We propose to investigate whether automatic computational approaches can be used to detect falsehoods in written text. Specifically whether certain features can be collected from reliable and non-reliable sources and whether they can be classified as such using existing models. We will also check for specific features of deceptive text. Conroy, Rubin and Chen (2015) mention several classification methods but note that simple n-grams and Parts-of-speech tagging

DJ ASCII – 18

6

do not take into consideration the context of the text unless used with other complex methodologies such as Probabilistic Context Free Grammars (PCFGs). We propose to implement the following commonly used classification models: 1) Recurrent Neural Network 2) Long Short-Term Memory 3) Random Forest Classifier 4) Logistic Regression The project analyses and compare their performance to determine the feasibility and success rate of an automatic fake news detection system based on the three techniques.

PAPER ID: AI16

Artistic Style Transfer Using Deep Learning Pankil Daru1, Siddhant Gada2, Meet Chheda3, Prof. Purva Raut4

1,2,3 Department of Information Technology

D. J. Sanghvi College of Engineering Mumbai.

E-mail:1 [email protected], [email protected], [email protected] [email protected]

Since past few years, Neural Style Transfer has become a gravitated topic both in academic literature and industrial applications. Neural Style Transfer is a concept where the style of an abstract artwork is superimposed on an image. Ever since its foundation it is receiving an increasing attention from computer vision researchers. Several ways are being proposed to either extend or improve the first algorithm proposed by Gatys et al. The algorithm could be applied to immense range of images and has fabricated assorted applications. We approached this concept by our own model based on Gatys on designing of drapes. This model permits a user to use his own content and style image. It provides user to enter multiple styles on a single content image. Also, a user can apply masking to provide the user with two different styles in the different regions of the content image. The output will provide an attractive image.

mailto:[email protected]

DJ ASCII – 18

7

PAPER ID: AI17

An Offline Handwriting Recognition Approach for Ruled

Pages using CNN and LSTM Jay ashok lal1, Meet mukadam2, Viral pasad3


K. J. Somaiya College of Engineering Mumbai, India


We present an end to end system for recognizing handwritten text from ruled pages. The proposed system utilizes the inherent line separation in ruled pages for segmentation and then uses Convolutional Neural Networks (CNN) and Bidirectional Long Short Term Memory Networks (BLSTM) for recognition of the same. The approach is also expected to work for unruled pages, after explicit line segmentation.

PAPER ID: AI19

Nutritional Estimation from Fast Food Images and

Alternative Recipe Suggestion for Diabetic Patients Kaiz Merchant1, Yash Pande2, Ashutosh Singh3, Prof.Mr. Pankaj Sonawane4

1,2,3,4 Department of Computer Engineering,



In recent years, obesity and health of diabetic patients have become major issues. To address the issue of obesity, it is very important for a person to know the amount of calories he or she is consuming, while for addressing the issues related to health of diabetic patients, sugar and carbohydrate levels are important. We propose a novel deep learning Convolutional Neural Network based system that can effectively run on smartphone devices, that not only provides the appropriate nutritional estimates but also suggests alternate food recipes for diabetic patients. Our system pre-processes the image as a first step in the process and then food recognition is carried out. After this, calorie, sugar and carbohydrate content is shown to the user and alternate healthy options for diabetic patients are presented. Our deep learning framework attained an accuracy of 96 percent for the 5 food categories we trained it for. By user experiments, effectiveness of the proposed system was confirmed. The future scope includes expanding to more food categories and optimising the application for better results.



DJ ASCII – 18

8

PAPER ID: AI20

Smart Health Monitoring Shweta Pujari1, Sneha Salunke2, Darshana Samala3, Nilofer Shaikhr4

1,2,3,4, Department of Electronics and Telecommunication,

SIES Graduate School of Technology Nerul, Navi Mumbai


This paper proposes to automate continuous monitoring of body parameters. We have designed an Internet of Things (IoT) based smart wearable system capable of monitoring pulse rate, temperature and sweat values(only emergency condition) using biosensors that are interfaced with PIC microcontroller. The recorded data is sent to the user via Bluetooth. An android App is developed to get access to the data which is being transmitted. The data is stored in the cloud so that the doctor and relative can access the data remotely.

PAPER ID: AI22

Autonomous Car Onkar Walavalkar1

, Suraj Parab2, Vikas Tiwari3

1,2,3, St. John College of Engineering and Management Village Vevoor, Manor Road, Palghar.


The idea of an autonomous car incorporates real-time detection of vehicles and other objects in the near vicinity and navigation without human intervention. Implementation of the same requires use of various Machine Learning algorithms such as Computer Vision and Reinforcement Learning for making superlative decisions for steering of the vehicle. The autonomous car proposed in this paper uses an intelligent system which is designed to perceive its environment using a real-time video feed and performs the different sets of actions for proper navigation. The system is initially trained on Darknet Neural Network, using Q-learning technique which is a type of Reinforcement Learning Algorithm. Once the system is trained, it will be capable of identifying objects present in its field of vision and performing the necessary driving decisions associated with the Q-tables. As a proof-of-concept, the system is implemented using miniature version of a car which will be trained for a specific terrain and based upon the training it will be tested under similar environments.

DJ ASCII – 18

9

PAPER ID: AI23

Citizen Service Problem (CSP) Aditya Kapadia1, Lucky Rathod2, Jigar Panchal3




In the recent monsoons, due to continued waterlogging, potholes wrenched havoc in the daily lives of people. From traffic jams to injuries caused simply due to walking through waterlogged roads, all this can primarily be attributed to improper garbage disposal and the pothole pandemic. If there were an easy means to tackle this issue, majority of the problem faced during heavy rains could be averted. Our application is specially designed to handle this large-scale garbage and pothole pandemic using our own custom built convolutional neural Network. An application that can do everything, from capture, to report and classify, to follow through, to tag on the map and even provide feedback.

PAPER ID: AI24

Automated Essay Generation Dhruvil Mehta1, Rajat Bhagat2, Aditya Dubey3



Student's ability to write essays on various different topics shows their command over the language. Students having knowledge about their domain cannot succeed as they are unable to express their ideas. To address this problem, we propose a system to generate factual essays. The applications of essay generation are endless. Our proposed system can be seen as a series of 5 distinct steps: Source Selection, Extractive Summarization, Knowledge Representation, Clustering and Ranking, and Natural Language Generation. We have considered Wikipedia as out input source. Information is extracted and represented in the form of different clusters. Abstractive summarization techniques are used to generate a unique set of sentences for the essay. The proposed system currently aims to generate factual essays pertaining to "countries" dataset. The system’s applications are far reaching, from educational, to corporate. It can be used to generate factual essays for the purpose of aiding students. It reduces the amount of information to be parsed, hence simplifying the task of processing said information for a human.

DJ ASCII – 18

10

PAPER ID: AI25

Real Time Surveillance System on Hadoop Image

Processing Interface Prof. Er. Mohammed Ahmed Abdul Mannan1, Shamuwel Ahmed Ansari2, Shebaz Ansari3, Sayed Adnan

Husain4


M.H.Saboo Siddik College of Engineering, Mumbai, India

Traditional security systems work to avoid crimes as much as possible. Real time Surveillance gives an opportunity to prevent any criminal activity before they can happen. Implementing security measures are also very complex and takes a lot of time and also requires human interference. An autonomous security system will make security economically feasible and it works quickly. Using facial, object and behavior recognition on the input feed provided by CCTV cameras, various criminal activities can be detected, and authorities will be assisted to take desired action. Covering large number of CCTV’s distributed over wide space can generate lots of data and requires tremendous processing power to process this data. Hence, we will use “Hadoop’s image processing interface (HIPI)” to distribute the processing task over the cloud network, so communication between authorities of various areas is enhanced.

PAPER ID: AI26

Voyageur- A Smart Trip Planner Akshen Kadakia 1

, Urvi Mistry2, Devanshi Desai3, Prof. Mitchell D’silva4

1,2,3,4 Department of Information Technology,

Dwarkadas J. Sanghvi College of Engineering Mumbai, India


In today’s busy world, travelling can be one of the therapies that can reduce all our stress. A traveller has lots of dreams but lack of plans. If the user is sure about the place to visit, he never finds all the required information at the same place. Even if he finds all the information of the destination; he is not sure about the reviews of the places in that destination and it would take him a number of days to make an itinerary. There are many existing systems such as tour planning websites, offline customized tour planners, etc. to help plan an itinerary, but they don’t suffice all the user requirements. Also, they do not provide instant customization. Our survey confirmed that the users face some problems with the existing systems. Thus, this paper presents a trip planning application ‘Voyageur’ which will help the users to plan their trips more efficiently. It will help the

mailto:[email protected],%204

mailto:[email protected],%204

DJ ASCII – 18

11

users to find all the information required at a single place. Also, it will give freedom to the user to build their customized itinerary automatically by taking into consideration various factors such as user preferences, the shortest distance between places to visit, etc. It will provide the user with many alternative plans for their trip using genetic algorithm. It will also suggest best hotels and places to visit based on ratings available on the internet. Thus, this instant itinerary generation will save a lot of user’s time.

PAPER ID: AI27

Smart Meter Data Compression and Pattern Extraction Ankit Shah1, Shweta Sunderkrishnan2, Dhrumil Dhutia3, Prof. Aruna Gawade4




Smart meters are technologically advanced IoT devices that record real-time consumption of electric energy and communicate that information, at least daily, back to the utility for monitoring and billing. On an average, the readings are taken every 15 minutes, thus the storage of this humungous yet valuable dataset has to be done at a reduced cost along with effective extraction of useful information. Electricity theft is one more issue that needs to be solved. Methods for stealing electricity can include tapping energy directly from an overhead distribution feeder, tampering meters. Thus with the advent of machine learning technologies, patterns in real-time energy consumption data can be analyzed to detect illegal consumers. By applying supervised learning techniques like SVM on customer consumption data, customers can be classified into illegal and genuine. Moreover, by taking into account various parameters like deviation in electricity consumption, season, date, time and ID of customers a more less accurate prediction can be made regarding electricity theft. Our paper aims to provide a solution for high storage cost issues by utilizing optimum compression algorithm like KSVD and sparse approximation algorithm OMP. Our paper also proposes use of machine learning algorithms to detect electricity meter tampering. Patterns from data will be extracted by applying classification techniques on customer energy consumption data to classify genuine and illegal customers.


DJ ASCII – 18

12

PAPER ID: AI29

Intelligent Alzheimer’s Detector using Deep Learning Mukul Puranik1, Himanshu Shah2, Keval Shah3



Researchers of the era are constantly striving to achieve accurate, precise algorithms incorporated in highly affordable models, trying to assist the medical practitioners in solving complex medical problems (Alzheimer’s). Deep Learning is state-of-the-art learning algorithm in classification and exceptionally efficacious in extracting high level features from multi-dimensional data. In this system we use Convolutional Neural Network particularly for classification of fMRI clinical data on stages of Alzheimer’s disease brain from Diseased (AD), EMCI to those who are normal, having healthy brains. Usage of MRI data has already been done [1] for binary classification. We aim to generalize the classifier into categorizing the images into three different distinct classes. The deep learning pipeline involved critical steps of correctly preprocessing 4D fMRI images i.e 3D images varying with time. The preprocessing steps involved, proved to be critical in having distinct 2D image slices of Normal, EMCI and AD for better accuracy in understanding the most discriminative features in the fMRI brain scans. Transfer Learning is the concept involved wherein we utilize pretrained complex deep models for classification of images. Advantage of this learning over constructing a new convolutional network is that the knowledge gained during training of ImageNet Dataset fastens the learning process in addition to increased accuracy. We have adopted Inception Resnet V2 model and hope to achieve a competitive accuracy. The aim of the project is to create an Alzheimer’s Detector ousting the accuracy of modern radiologists so as to reduce the effort and money of consulting a Radiologist.

DJ ASCII – 18

13

PAPER ID: AI30

Design and Implementation of Neuromuscular

Stimulation to Bypass Nerves System Amit Katariya1, Shantanu Madiwale2, Meet Mehta3




Paralyzed person or the person with physical disorders won’t be able to move their body parts viz. hand or arm because of abnormal behavior of spinal cord or its injuries. In such cases, the traditional solution is to use neuroprosthetic arm replacing paralyzed body part. This kind of surgery basically involves an artificial arm or robotic arm which is controlled by brain which is expensive also. This proposed research work will avoid the need of replacing the human body part. Instead of using prosthetic arm, paralyzed person can use proposed nervous bypass system which will create an external bridge to control paralyzed hand without any surgery. This interaction through machine will make use of embedded devices to provide the solution to surgery and this interaction can be happen by another human to human like normal person can also control hand of Paralyzed person, this can be happening wirelessly also.

PAPER ID: AI31

Human Understanding Analyzer (Machine Learning) Manogya Prasad1, Rizwan Japanwala2, Harshil Vora3, Prof. Mrs. Lakshmi Kurup4




The aim of the project is to create an AI based application that can analyse emotions in frames from a live video feed of a person learning and determine the relative understanding level of the person during and after the learning process. This is achieved by using a Convolutional Neural Network trained by the FER-2013 Dataset “Emotions In The Wild” for emotion recognition and an algorithm to determine relative level of understanding using that information. Output will finally show the relative variation in the person’s level of understanding over time. The patterns in the variation can then be easily interpreted to determine whether the person is understanding and at what point of time during the learning process did it drop or increase. A web based tool for analysis of understanding level of the person is used at the end of session.


DJ ASCII – 18

14

PAPER ID: AI32

Heart Rate Evaluation and Risk calculation Chintan Devda1, Jaydeep Gami2, Meet K Mehta3



Artificial Intelligence techniques have been successfully employed in disease diagnosis, disease risk evaluation, patient monitoring, robotic handling of surgeries and predicting effect of new medicines. This paper proposes and evaluates Genetic Neuro Fuzzy System for diagnosing Hypertension risk. Risk factors viz. Systolic and Diastolic Blood Pressure, Body Mass Index, Heart Rate, Cholesterol, Glucose, Blood Urea, Creatinine and Uric Acid have been taken as inputs to the system. The system classifies the input samples into Low, Medium and High risk samples. The fuzzy logic qualitative approach is integrated with genetic algorithm to diagnose the presence of the disease.

PAPER ID: AI33

Text to Image Generation using Deep Learning Sudhir Bagul1, Kumpal Dhruv2, Nikhil Kamat3, Pooja Kulkarni4



This paper reviews the methods to generate images for a given text caption. The process works by converting the given text caption sequences into text embeddings, using skip-thought vectors or by using char-CNN-RNN text encoders conditioned on the images and text captions of the used dataset. The text embeddings are then passed to a generative convolutional neural network which outputs the generated images, which are similar but not included in the dataset used. This generative network is trained by the process of adversarial training, implemented in Generative Adversarial Networks.

DJ ASCII – 18

15

PAPER ID: AI34

Calamity Evacuation using Social Network Parth Vora1, Mili Desai2, Aditya Vallat3



Calamities happen from time to time. Disaster Management Teams’ focus is to send rescue and relief teams on site at the earliest so that the affected persons can be evacuated and provide the required care. Responding to a calamity / disaster event quickly and efficiently can mean a matter of life and death for some affected persons. In this project, we plan to build a system that would provide emergency personnel the ability to collect information in real time about disaster events, track the resources (rescue teams and helpers) and manage them. It allows the head responders, rescue teams and helpers to manage multiple incidents simultaneously. With the help of apps built for an android device, the head responder will be able to quickly and most optimally assign rescue teams to different locations and rescue teams will get information on the priority and order in which they need to rescue the affected persons. This project also describes the implementations of commander app, responder app and victim app and the algorithms used to come up with an effective and optimum evacuation strategy. We also have implemented algorithms to optimize the total time taken to evacuate a calamity affected area. Along with it the commander can review the entire process once it is done and then provide feedback in the application itself. This will help in the future use of this application.

DJ ASCII – 18

16

PAPER ID: AI38

IoT Based Intelligent Medical System-IMED Yashraj Kotian1, Pawankumar Dubey2, Romil Badami3, Vivek Giri4, Dr. Rekha Sharma5

1,2,3,4,5, Thakur College of Engineering and Technology, Kandivali- East, Mumbai, India

E-mail: 1 [email protected], 2 [email protected], 3 [email protected] 4 [email protected], 5 [email protected]

These days healthcare is a big factor for systems which are struggling with aging population, prevalence of chronic diseases, and the accompanying rising costs. In response to these challenges, researchers have been actively seeking for innovative solutions and new technologies that could improve the quality of patient care meanwhile reduce the cost of care through early detection/intervention and more effective disease/patient management. It is important that the upcoming system should be preventive, predictive, personalized, participatory, patient-centered, and precise, i.e., patient-health system. Health informatics a rising interdisciplinary area to increase efficiency of patient-health, mainly deals with the gathering, transferring, processing, storage, finding, and use of different types of health and biomedical information. The two main important technologies of health information are sensing and imaging. This system focuses only on sensing technologies and the recent developments in sensing devices for continuous health monitoring and accessing the information.

PAPER ID: AI42

On Panel Auxiliary Warning System for Indian Railways Saurabh S. Deone1, Harshal D. Waje2, Pravin Kanase3

Saraswati College of Engineering, Kharghar, Navi Mumbai


Considering the current state of turmoil in the Indian railways, the controversy surrounding the

use or rather the ineffective use of technology in the railways, there is an aggravating need for

some improvement in technology and at a low cost. One quick-fix solution to this is the

indigenously developed, technologically sound and readily available ‘Anti-collision Device’. But

the drawback of that technology is its high cost of installation. This drawback is overcome by the

development of an ‘On Panel Signaling System in trains with Auxiliary Warning System (AWS)’.

The main aim of this project is to develop a panel which can be mounted on the train’s dashboard

so that the motorman has the present status of the nearest signal at his disposal. If the motorman

still overlooks the signal, the train will be brought to a halt .This is achieved using a circuit made

up of micro-controllers, radiofrequency & infrared frequency transmitters and receivers and

RF encoder and decoder.



DJ ASCII – 18

17

PAPER ID: AI43

Real Time Drowsiness Detection System Mr. Shailesh Sangle1, Bharat Rathore2, Rishabh Rathod3, Aakashkumar Yadav4,

Abhishek Yadav5

1,2,3,4,5 Department of Computer Engineering

Thakur College of Engineering & Technology, Mumbai.

The proposed paper describes how efficiently a non-intrusive computer vision based concept be used for detecting drowsiness of the Driver. A System will be developed which aims to improve the road safety using advanced technology. The proposed system will use a basic Web-cam interfaced with Raspberry-Pi that points directly towards the driver’s face and monitors the driver’s eyes and Face in order to detect fatigue. If Symptoms of Fatigue such as closed eyes or head lowering occurs a warning signal is issued to alert the driver. The product prototype proposed is unique to the road safety purpose. It uses the concept of image processing. Open-CV will be integrated with the python and deployed on raspberry-pi. Harr classifier algorithm used will help to determine if the eyes are open or closed. The algorithm developed is unique to any currently published papers in terms of its application with Raspberry pi, the primary objective of the project was to make the system portable and hence can be applied in existing vehicles to improve the safety by giving driver a feedback when he feels drowsy. The algorithm uses image binarization to mark the edges of the face. Once the face area is found, the computation of the horizontal averages in the area helps us locating the eyes. Eye regions in the face shows great intensity changes, the eyes are located by facial landmarks detector file which is in dlib library. Once the eyes are located, distance is measured between the intensity changes in the eye area which helps us to determine whether the eyes are open or closed. Large distance computed corresponds to eye closure. If the eyes are found closed for few consecutive frames, the system draws the conclusion that the driver is falling asleep and issues a warning signal. There is also a provision of monitoring the head region in the frame, similarly if for few consecutive frames if the head region is out of the frame also triggers a warning condition. The system is also able to detect when the eyes cannot be found and work under reasonable lighting conditions.

DJ ASCII – 18

18

PAPER ID: AI44

Triple Technique Diagnosis Using Machine Learned

Classifiers Het Sheth1, Dhruman Shah2, Mihir Gada3, Purva Raut4

1, 2, 3 Department of Information Technology

D. J. Sanghvi College of Engineering, Vile Parle, Mumbai, India

E-mail: 1 [email protected], 2 [email protected], [email protected]

In the present scenario, breast cancer has become one of the most common diseases amongst women. Despite this, not all public hospitals have the facilities to diagnose breast cancer through the Triple Technique (Palpation, Mammograms and FNAC). Delaying the diagnosis may increase the chances of a cancer to spread throughout the body. Machine learning techniques have been benevolent in the detection and diagnosis of various diseases due to their accurate prediction performance. Various classifiers may provide differently desired accuracies and it is, therefore, exigent to use the most fitting classifier which provides the best accuracy. This paper documents a study of 3 machine learned classifiers, namely, Support Vector Machines (SVM), Logistic Regression (LR) and K-Nearest Neighbors (KNN) on the Wisconsin Diagnosis Breast Cancer (WDBC) dataset. The performance of these algorithms will be analyzed using their classification accuracy and a confusion matrix. We will also implement Singular Value Decomposition (SVD) and analyze the results by comparing them.

PAPER ID: AI45

On Road Activity Monitoring System Tirth Patel1, Viral Shah2, Vipul Hodge3, Raj Desai4

1,2,3,4 B. Tech Computer Engineering

K.J. Somaiya College of Engineering, Mumbai, India


GPS and GPRS have revolutionized our lives with many real time applications which are being

used to make our lives much better and faster at the same time. This paper introduces a new

system for accident detection as well as reporting and is based on our BE project. Our developed

system, On Road Activity Monitoring System (ORAMS) is mainly designed with the aim of

reporting users within the proximity of an accident location so that necessary actions can be taken

as soon as the notification is received which in turn can save a precious human life. Also we’ve

added nearby places for a person within the accident vicinity or proximity so that these can be

used as and when required by the people using the android application in case of such an

emergency. The main aim is to notify users that an accident has occurred within their vicinity so

that the accident victim can be given medical attention as soon as possible by the one’s receiving

the notification.

DJ ASCII – 18

19

PAPER ID: AI49

Solar Energy prediction using Artificial Neural Network Pranali R. Mane1, Vivek H. Patel2, Saurav S. Yadav3, Prof. Mr. Abhijit R.Joshi4




The recent increase in erratic global climatic conditions such as global warming is alarming, and one needs to switch to natural and renewable sources of energy like solar and wind. With respect to solar energy, industry experts and stakeholders should create awareness amongst customers to adopt this eco-friendly solution. Our product, OurSolar, aims to revolutionise the solar industry by giving real time Direct Normal Irradiance (DNI) predictions for the future, which is the amount of solar radiation received per unit area by a surface that is always held perpendicular (or normal) to the rays that come in a straight line from the direction of the sun at its current position in the sky [1]. The solution incorporates the varying factors of cloud cover, sunlight duration, average temperature, precipitation, vapour pressure, atmospheric pressure, humidity and altitude to give an almost near-to-real DNI value. Apart from helping companies sell their solar solutions, OurSolar will also suggest the customers a suitable solar system depending on his location and requirements, thus helping him with his energy plans. For time being, the product spans over the state of Maharashtra covering all the 36 districts and an area of 307,713 sq km. For prediction, an artificial neural network model is used with 100 hidden layer neurons which gives an accuracy of 92.63%.

DJ ASCII – 18

20

PAPER ID: AI51

Audio Recording and Analysis to Detect and React to

Anthropogenic Disasters Shalin Parikh1, Raunak Vijan2, Zaid Merchant3, Prof. Mr. Pratik Kanani



The working process of our proposed system is explained as: Once the incident is detected via audio content retrieval, an alarm can be immediately issued to the local police officers and immediate intervention can be performed to contain the event. The fundamental idea is to create a taxonomy to organize unstructured hazard sound data. The hierarchical formation can significantly facilitate browse, search and classification of acoustic patterns. In order to characterize predominant patterns in emergency sounds, unsupervised acoustic feature learning algorithms are employed. The methods can effectively extract effective features that are invariant to background noise. On creation of disaster taxonomy, we introduce probabilistic distance metrics in both Euclidean and Grassmannian spaces to quantize difference between hazard sound categories. A taxonomy can be subsequently built using well-defined categorical distance measures in agglomerative fashion. At multi-class emergency sound recognition stage, we devise methods to embed hierarchical dependencies in acoustic data in classification algorithm. We intend to implement using a test data set for the training set to classify all categories. Actual implementation can be proven by using a microphone attached to a Raspberry pi for the IoT aspect of our project.

DJ ASCII – 18

21

PAPER ID: AI52

Identification of Potential Cyberbullying Tweets using

Hybrid Approach in Sentiment Analysis Akankshi Mody1, Reeya Pimple2 , Shreni Shah3, Dr. (Prof.) Narendra Shekokar4

1,2,3, Computer Department, D. J. Sanghvi College of Engineering Mumbai.

4 HoD, Computer Department, D. J. Sanghvi College of Engineering Mumbai

E-mail:1 [email protected], [email protected], [email protected], [email protected]

As the Internet is increasingly used by individuals to express their opinions, cyberbullying is becoming predominant. Research has shown that cyberbullying results into serious consequences. The aim of this study is to implement an automated cyberbullying detection method that identifies cyberbullying in Tweets. We propose a cyberbullying detection technique that uses a hybrid approach of analyzing tweets. We first segregate the textual content of the tweet from the emoticons and then perform sentiment analysis on both these parts to identify the emotion and intent of the tweet. The text sentiment analysis involves a lexicon based approach using SentiWordNet and a machine learning approach which reinforces the result obtained from the SentiWordNet. The emoticon sentiment analysis involves a lexicon based approach. We assign an overall polarity to each Tweet which will classify it as a negative or positive tweet.

PAPER ID: AI57

Application of ML Techniques for the Analysis of

Hypertension and Prediction of Vein Function in

Hemodialysis Dr. Gresha Bhatia1, Mihir Wagle2, Neeraj Jethnani3, Juhi Bhagtani4, Aishwarya Chandak5

1,2,3,4,5 CMPN Department, V.E.S.I.T Chembur, Mumbai, India

Millions of patients worldwide suffer from Kidney failure and require dialysis. In most cases, dialysis is started after the kidney function of the patient falls below a threshold. In this scenario the patient’s kidney is essentially nonfunctional. In order to conduct dialysis, native arteriovenous fistulas are constructed to increase blood flow in the superficial vein, and hence facilitate dialysis. Over time, as dialysis continues, the patient may suffer from hypertension and reduced vein function leading to the collapse of the fistula. The ultrasound doppler test for checking the state of the fistula is expensive and doing it again and again is not feasible. The proposed work explores the Chronic Kidney Diseases and proposes a mechanism that uses optimised data points to predict health of fistula.



DJ ASCII – 18

22

PAPER ID: AI59

Prediction of Personality based on Handwriting Sharanya Ojha1, Abhishek Shah2, Sagar Shrinagarpure3

1,2,3, Student Department of Engineering,

D. J. Sanghvi College of Engineering Mumbai Organizations recruiting people are constantly on the lookout for people with specific qualities

and traits. We felt the need for a method whereby organizations can properly analyze people

based on their personality type. Handwriting conveys important information about the physical,

mental and emotional state of the writer during writing and also about his overall behavior or

personality traits. In this project, we develop a system which can analyze such a handwritten

sample and predict the personality of a person among the 16 Myers Briggs personality types.

With the advancement in fields such as Machine Learning and Image Processing, the analysis of

handwriting can be automated completely. This paper proposes to create a classifier using SVM

after processing a scanned sheet of sample handwriting to reveal the personality of a person.

PAPER ID: AI60

Multiclass Classification of Imbalanced DataStream Pratik Parekh1, Medha Shah2 , Utsav Shah3

1,2,3 Computer Department, D. J. Sanghvi College of Engineering Mumbai Imbalance data stream classification deals with data streams having very skewed class distributions. Seven vital areas of research in this topic are identified, covering the full spectrum of learning from imbalanced data: classification, regression, clustering, data streams, big data analytics and applications. Our preliminary focus out of these seven areas is on the continuous and skewed data stream. The stream processing frameworks that are designed to process the streaming data arrives in real time. Multi-class imbalanced classification is not as well-developed as its binary counterpart. Here we deal with a more complicated situation, as the relations among the classes are no longer obvious. A class may be a majority one when it is compared to some other classes, but a minority or well-balanced for the rest of them. When dealing with multi-class imbalanced data we may easily lose performance on one class while trying to gain it on another. A deeper insight into the nature of the class imbalance problem is needed, as one should know in what domains does class imbalance most hinder the performance of standard multi-class classifiers when designing a method tailored for this problem. In this project, a framework for modelling and classifying the streaming data when the classes of the data samples are imbalanced were proposed. Modelling methods implemented include techniques such as data pre-processing, machine learning algorithms and model evaluation. The combined challenges posed by multi-class imbalance and online learning, are resolved with a more effective and adaptive solution for the classification of this data.

DJ ASCII – 18

23

PAPER ID: AI61

Dermatological Disease Classification Sagar Dedhia1, Pratik Jayarao2 , Yash Doshi3

1,2,3 Department of Computer Engineering

D. J. Sanghvi College of Engineering Mumbai E-mail:1 [email protected],2 [email protected],3 [email protected]

Skin diseases are amongst the most common health issues around the globe. Identification of skin diseases requires a high level of expertise as there are various visual aspects which have to be taken into consideration. In this project we propose an approach that uses IP feature extraction along with deep neural networks to classify various kinds of dermatological skin diseases. We have performed feature extraction using different types of image processing techniques. With help of artificial neural network and convolutional neural network we have classified 6 different skin abnormalities.

PAPER ID: AI64

Machine Learning Approach to Foretell the Probability of a

Crop Contracting a Disease Viraj Mehta1, Chahat Jain2, Karan Kanchan3

1,2,3, Department of Information Technology

Dwarkadas J. Sanghvi College of Engineering, Mumbai

E-mail:1 [email protected],2 [email protected],3 [email protected]

India has an agricultural economy; implying it is highly dependent on the sustainable yield of the crops. Unfortunately, the production of crops has decreased by 15-25 % over the last few years with crop diseases being the main culprit. The diminished crop yield and productivity has led to a decline in the country’s economy and consequently, the farmer’s community has been profoundly affected. In this project we have targeted the environmental vertices of the Disease triangle of crop, which include various parameters like weather and soil conditions of the land taken into consideration. It will output the probability prediction about the diseases that a crop can contract on the basis of the weather and soil vectors. Here, we have taken four Machine Learning Algorithms – Decision Tree, Multi-Output Regress or, Multi-layer Perceptron and Random forest to predict the probabilities for various diseases the crop can contract and compared them to get the best prediction. These prediction outputs by the system will assist the farmer in taking appropriate disease prevention techniques and increase the productivity and quality of their yield.

mailto:[email protected],2




DJ ASCII – 18

24

PAPER ID: AI65

Medical Intelligent Record Assistant Suhail Barot1, Dhanashree Jatar2

1,2 Department of Computer Engineering,

D. J. Sanghvi College of Engineering Mumbai

In the healthcare industry, natural language processing has many potential applications. NLP can enhance the completeness and accuracy of electronic health records by translating free text into standardized data. It may be able to make documentation requirements easier by allowing providers to dictate their notes, or generate tailored educational materials for patients ready for discharge. However, one problem strikes us as more pressing than the rest. Today, hospitals are, sadly, one of the busiest places to work. With the burden of treating 50+ patients every day, doctors are unable to properly document each case. Therefore, there is a lot of data loss. Many a time, this data can be the difference between life and death for a patient. We envision a tool that listens to, understands and organizes diagnosis related conversations between doctors and patients. It also uses a specialized semantic network known as a “Knowledge Graph” to build contextual relationships between people, diseases, and symptoms. We call her the Medical Intelligent Record Assistant, or just MIRA. MIRA processes speech, converts it to text, and extracts contextual information from the dialogue. This includes symptoms, diseases, medicines and other medical jargon that is essential to keep track of the patient’s progress. It uses this information to build a knowledge graph, which allows doctors to use it as a Decision Support System as well. An electronic health record (EHR), or electronic medical record (EMR), refers to the systematized collection of patient and population electronically-stored health information in a digital format. The raw information collected and processed by MIRA will be stored in these EHRs (along with the knowledge graph), which increases accessibility and portability. The graph based storage predicts possible diseases, related symptoms and other vital information to assist the doctor.

DJ ASCII – 18

25

PAPER ID: AI67

Automated Evaluation of Subject Answers using Text

Processing and Sentence Similarity Aditya Maniar1, Devansh Gada2, Chantelle D’Silva3

1,2,3 Student Department of Computer Engineering


Trends of online objective examination are already available but many courses require assessment in traditional way so that the subject understanding of the candidate can be evaluated which requires subjective assessment i.e. descriptive based examination. Thus in our project we are focusing on the inference process required for development of such type of systems. Our system will be able to assess one/many words and one (few) sentence(s) based answer with more than average efficiency. While answering single sentence answers paraphrasing is consider for assessing the variations occurring due to the use of vocabulary

PAPER ID: AI68

Game Automation Using Reinforcement Learning Bhargav Mehta1, Rohan Madhani2, Viral Lakhani3

1,2,3 Student Department of Computer Engineering,


E-mail: 1 [email protected] , [email protected], 3 [email protected]

The project aims to replicate a system that combines deep learning methods and reinforcement learning in order to create a system that is able to learn how to play games on its own. A model free reinforcement methodology named ‘Q-Learning’ is used. Therefore, the core algorithm of Deep Reinforcement Learning used is ‘DQN’. DQN is combined with Reinforcement Learning (RL)—a machine learning framework that prescribes how agents should act in an environment in order to maximize future cumulative reward (e.g., a game score). Foremost among these was a neurobiological inspired mechanism, termed ‘experience replay’, whereby during the learning phase DQN was trained on samples drawn from a pool of stored episodes. The agent is not given information about game - it must learn these representations and directly use the input and score to develop an optimal strategy. The system has access only to the visual information i.e. the screen of the game and the scores. Based on these two inputs the system learns to understand which moves are good and which are bad depending on the situation on the screen.



DJ ASCII – 18

26

PAPER ID: AI70

Autonomous Driving Car using Neural Networks Raj Chaudhari1, Shivani Dubey2, Jayesh Kathale3

1,2,3 student Vidyalankar Institute of Technology Wadala(E), Mumbai 400 037


In this rapidly moving world, cars and other modes of On-road vehicles play a very important role. According to a study, it was approximated that approximately 48% of world’s mobility depends upon On-road vehicles. Such huge traffic is also responsible for a humongous number of deaths and injuries because of accidents. According to the study conducted by Road-safety committee, 410 lives are lost every day in India because of accidents. A large proportion of reason for these accidents is driver negligence, ignorance of traffic rules and other driver-related mistakes. An Autonomous Driving Car is a possible viable solution to this which help in bringing down the fatalities to a minimal extent.

PAPER ID: AI74

Image Caption Generation Using Deep Neural Networks Siddhant Agarwal1, Himani Deulkar2

1,2, Student, Department of Computer Engineering

Vidyalankar Institute of Technology, Mumbai

E-mail: 1 [email protected] , 2 [email protected]

Captioning an image is a problem of generating a textual description given an image as input. It is a challenging problem in Artificial Intelligence which requires techniques from computer vision to interpret the contents of the image and techniques from natural language processing to generate the textual description. The model uses Convolutional neural networks to extract features from an image, these features are then fed into a Long Short-Term Memory (LSTM) to generate a description of the image in English. The text captions are then converted into speech which can assist visually impaired people to get a better sense of the environment.


DJ ASCII – 18

27

PAPER ID: AI76

Simulation of Self-Driving Car in a Simulated Environment Atharva Kunte1, Avasyu Bhatia2, Ziyad Dhuka3




Our system is an implementation of driverless cars in a simulated environment such as a car driving simulator. The chief reason a simulated environment was chosen over real world is to mitigate the risk to human life. Also, a simulated environment would reduce the material losses as well as the overall cost of the implementation process. Once the system has attained a high proficiency in the simulated environment, the system can be migrated and perfected for the real world with some modifications. The software will take the first person view of the car and feed the image to our system. We train our system to map raw pixels from a single front-facing camera directly to steering commands in our simulation.

PAPER ID: AI78

Safe Driving System Abhishek Shukla1, Yeshwant Ranka2, Bhavya Shah3, Prof.Mrs. Sindhu Nair4

1,2,3, Student, 4 Assistant Prof. Department of Computer Engineering,


E-mail: - 1 shukla1996abhishek @gmail.com, 2 yeshwant. ranka1010 @gmail.com, 3bhavyashah9 @gmail.com

Accident prediction is one of the most critical aspects of road safety, whereby an accident can be predicted before it actually occurs and precautionary measures taken to avoid it. For this purpose, accident prediction models are required in road safety analysis. Artificial intelligence (AI) is used in many real world applications, especially where outcomes and data are not same all the time and are influenced by occurrence of random changes. This paper presents a model for Safe driving system. AI techniques are surveyed for the detection of unsafe driving style and crash prediction. A number of statistics and data set are used to predict the accidents by using different vehicle and driving features are also covered in this paper. The parameters used are in terms of datasets and prediction performance. We also provide a list of datasets and simulators available for the scientific community to conduct research in the subject domain.

DJ ASCII – 18

28

PAPER ID: AI79

NaYaNa – A Phonetic script to Make Communication Easy Prachi Rahurkar1,Deepa Ramrakhiani2, Smriti Rao3

1,2,3, Student, Thadomal Shahani Engineering College E-mail: [email protected], [email protected], [email protected]

The purpose of this research is to propose an alternate character set of IPA. This script called 'NaYaNa' is developed to make communication easier by making reading and writing dyslexia-friendly. In English and some other languages, the rotation of one alphabet leads to another alphabet. For example, when 'N' is rotated, we get „Z‟. This creates confusion for dyslexic students. NaYaNa omits this ambiguity as the characters are designed taking under consideration unique rotational and mirror symmetry. That is, if „d‟ is used, ‟p‟ and „b‟ will not be used for any other phoneme. Similarly, each of the following pairs of characters, „N‟ and „Z‟, „M‟ and „W‟, „n‟ and „u‟ etc. would represent the same phonemes if they were used in NaYaNa. The shape of the character in NaYaNa has some metaphorical similarity to the shape and form of the vocal apparatus (lips, tongue, teeth, palate, cheeks and buccal cavity). The shape also corresponds with the quality of sound of the phoneme; such as sharp sounds have sharp edged characters, open sounds have open characters etc. The shapes may resemble some commonly used characters from other languages, but do not necessarily represent similar sounds. A one-to-one mapping of the characters is made with IPA as per the need. Since IPA has a defined Unicode space, this ensures that any designed font for NaYaNa can be used on any Unicode compliant computer. This script does not differentiate upper case and lower case and contains defined floor and roof modifiers on the unique characters. Also, it is designed to be visually accessible to human beings primarily. However, it is elegantly accessible for machines for text to speech (due to IPA compatibility) and OCR, due to the unique and unambiguous shapes. By identifying NaYaNa as a universal language, we will eliminate the difficulties that all dyslexic people face. This project will likely impact the future of communication among humans.

DJ ASCII – 18

29

PAPER ID: AI80

Context Based Question Answering System Avais Pagarkar1, Rudresh Panchal2, Swapneel Mehta3, Lakshmi Kurup4

1,2,3, Student, 4 Assistant Prof, Department of Computer Engineering,


Processing and understanding very large documents has always been a problem for humans. This is especially difficult in case of documents from highly technical domains like medicine and law. An interface which responds in natural language to a Human’s queries on any given document is the ideal solution. Such Question Answering systems, given the Document as context, parse it, understand it and answer questions based on the same. Our system consists of two major modules, the Document Retrieval and the Machine Comprehension system. For the Document extraction, we implemented a custom TF-IDF with intermediate weight hence processing documents faster. We use Deep Learning based approaches, namely a combination of Dynamic Co-Attention Networks and Bi-Directional Attention Flow models for the machine comprehension.

PAPER ID: CO07

ALLIO-All in One Multithreaded Server Darsh Shah 1, Vaibhav Shah 2, Sumit Busa 3

1,2,3, Department of Information Technology


E-mail: [email protected], 2 vshah2212@gmail, [email protected]

To develop a web application is attiring task. Number of problems are faced by the developers such as storing of scripts according to the server used. They have to also run the server manually after storing the script in respective folder. SQL to NoSQL data conversion also is a time consuming task. Also if developer prefers to use NoSQL for the application such as MongoDB, there is no proper user interface available for it, thereby leaving the developer with no alternative but to have in depth knowledge of the NoSQL database being used. Our paper deals which a proposed solution which aims to solve the above problems.

DJ ASCII – 18

30

PAPER ID: CO09

Security and Privacy for Data on Cloud using Image

Sequencing and Multiple Encryption

Zambare Jayesh Chhabildas1, Bedekar Ganesh Ramchandra2

1, 2 Department of Computer Engineering

Dilkap College of Engineering & Management Studies Neral 410 101, India

E-mail: 1 [email protected], 2 [email protected]

Cloud computing is emerging technology that enables access to system resources and applications. It may trace back to time when computing resources and applications were remotely time-shared from computer systems. Nowadays, applications are delivered in the internet cloud. In cloud computing the information is shared between end users. Due to the increasing demand for more cloud there is a security threats in cloud computing environment. Since cloud computing stores the data and resources in cloud environment, security has become the major issue in the improvement and reliability of cloud environment. Hence, security and privacy are the main challenges in cloud computing. In this paper, we are working on the security of cloud computing. To achieve security in cloud computing, we are going to use AES algorithm with RSA algorithm i.e. Multiple Encryption and to maintain the privacy, we have purpose an image-based password authentication. A password consists of one click-point per image for a sequence of images. The images can be user defined or predefined. Thus, image-based password provides usability and security.

PAPER ID: DM03

Lifestyle based Disease Prediction System Harsh Jain 1, Jay Jain2, Praful Kothari3, Prof. (Mrs.) Chetashri Bhadane4

1,2,3,4, Department of Computer Engineering,



Data Mining is has become one of the most important computing domains in today’s information age, from processing large amounts of data to finding relevant patterns. In present period, these advanced techniques are the epitome of establishing relevancy and context to elaborately describe clinical/medical and lifestyle data. Disease Prediction plays an important role in data mining. In order to analyze and generate relationships amongst various attributes, extensive Data Mining techniques and application specific algorithms are developed. The attributes which are used as input for our proposed model are age, gender, blood sugar, obesity level, cholesterol, level of smoking, exercise level, stress level and many more. These data are summed from different sources. Then it is assembled, integrated and cleaned up. After this step it is being able to predict the disease the user might possess. This will help individuals to made decision regarding their health and whether the lifestyle which they are adapted to is positive or not.

DJ ASCII – 18

31

PAPER ID: DM08

Data Preprocessing for Efficient Sentimental Analysis Shreyas Wankhede 1, Ranjit Patil 2, Sagar Sonawane 3


VIVA Institute of Technology Mumbai, India

E-mail: 1 [email protected], [email protected], [email protected]

Sentiment analysis is an important research area that identifies the people’s sentiments, opinions and emotions underlying a text. As the use of social media is increasing day by day, it plays an essential role in communication through technology. Twitter, which is one of the popular and largely used social media platforms for communication has more than 200 million tweets per day. Tweets are short in length and due to limited size of tweets people generally commit some mistakes while tweeting so pre-processing is necessary. The use of modern emoticons which are known as emojis that is largely used in social media communications that conveys variety of emotions. The purpose of this paper is to use N-gram method and Hidden Markov Model for Spell-Checking and Correction of tweets and also Emoji Sentiment Ranking method which is used to evaluate sentiment mapping of emojis by using sentiment polarity such as negative, neutral, or positive.

PAPER ID: DM10

Personalized Travel Sequence Recommendation on Multi-

Source Big Social Media Tarun Reddy 1, Jenila Sanghvi2, Deval Vora 3



This project is named as “Personalized Travel Sequence Recommendation” as it recommends the travel sequence using travelogues and community contributed photos along with heterogeneous metadata (e.g., tags, geo-location, and date taken) associated with these photos. Our approach is not only personalized to user’s travel interest but also able to recommend a travel sequence rather than individual Points of Interest (POIs) which most existing travel recommendation approaches fail to provide. We use topical package space which includes representative tags, cost distribution, visiting time and visiting season of each topic which is then mined to bridge the gap between user travel preference and travel routes. We utilize two kinds of social media viz. travelogue and community contributed photos. In order to get user topical package model and route topical package model, we map textual descriptions of user and routes to the topical package space. Famous routes are given a rank according to the similarity between user package and route package in order to recommend a personalized point of interest sequence to the logged in user. Then a route is selected from the top ranked routes to further optimize it according to social similar users’ travel records. Representative images with viewpoint and seasonal diversity



DJ ASCII – 18

32

of POIs are shown to offer a more comprehensive impression. We assess our recommendation system on the basis of 6 million Flickr images uploaded by ample number of users and approximately 20,000 travelogues covering 850 travel POIs in five famous cities, and show its effectiveness

PAPER ID: DM11

Anomaly Detection in Legal Documents Partik Aher 1, Kunal Doshi 2, Tushar Dey3, Prof. (Mrs.) Purva Raut4

1,2,3,4, Department of Computer Engineering,


E-mail: [email protected] 3 [email protected]

Legal documents have always been lengthy and it is difficult to read and understand them

completely. We devise a system that takes in document and highlights anomalies in the text. To

our knowledge, some categories of legal documents contain duplicated information that do not

require our attention. However, manually extracting nonduplicate information from documents

requires considerable amount of effort. Thus, we want to use machine learning algorithms to pick

up unordinary sentences for us. In our paper, we propose a set of algorithms that filters out

duplicate information and returns useful information to the user. Hence training a model using

semi supervised algorithms to fulfil our purpose to make a document shorter and less redundant.

Firstly, our input has taken from various sources and their public legal documents. This is cleaned

by using process like porter stemming. Then by LDA, we find the set of common words with very

frequency. These words are removed. After that, Word2Vec is used to convert this words into set

of vectors.

DJ ASCII – 18

33

PAPER ID: DM12

Understanding Short Texts through Semantic Enrichment

and Hashing

Devashree Patil1, Nitesh Jadhav2, Kanan Muthe3

Yadavrao Tasgaonkar College of Engineering and Management Chandai,karjat

E-mail: 1 [email protected], 2 [email protected], 3kananmuthe195.km @gmail.com

Clustering short texts (such as news titles) by their meaning is a challenging task. The semantic hashing approach encodes the meaning of a text into a compact binary code. Thus, to tell if two texts have similar meanings, we only need to check if they have similar codes. The encoding is created by a deep neural network, which is trained on texts represented by word-count vectors (bag-of-word representation). Unfortunately, for short texts such as search queries, tweets, or news titles, such representations are insufficient to capture the underlying semantics. To cluster short texts by their meanings, we propose to add more semantic signals to short texts. Specifically, for each term in a short text, we obtain its concepts and co-occurring terms from a probabilistic knowledge base to enrich the short text. Furthermore, we introduce a simplified deep learning network consisting of a 3-layer stacked auto-encoders for semantic hashing. Comprehensive experiments show that, with more semantic signals, our simplified deep learning model is able to capture the semantics of short texts, which enables a variety of applications including short text retrieval, classification, and general purpose text processing.

PAPER ID: DM14

FOREX Nimit Parekh 1, Manav Shah 2 , Varun Shah 3



The Foreign Exchange Market is the most advanced and liquid Financial Market. In a dynamic environment where variables are constantly changing, there is definitely the tendency of human’s error as well as human psychology affecting decision making of micro investors and resulting in losses. Our project thus steps in as an intelligent financial tool, to maximize returns and to minimize the leveraged risks taken on by investors. The financial model tool takes multiple minute to minute currency pair data as an input. With the help of multiple dynamic technical indicator variables, the tool easily identifies the trend along with the strength as well as the possibility of the trend. This dynamic data is then trains and tested on the system with the help of models i.e. Multiple Regression Model & The Support Vector Machine Model, they work in conjunction as classifieds of supervised learning to maximize the probability of a positive rate of return. This enables the investors and traders to make decision in an objective manner and thus serves as a precautionary and essential tool to maximize gains and minimize risk.

DJ ASCII – 18

34

PAPER ID: DM16

Document Clustering Using Ontology Maithili Shah1, Parth Mehta2

1,2, Department of Information Technology,


E-mail: [email protected], [email protected]

Nowadays, with the need to properly allocate the increasing number of research projects for peer review, there is a need to develop a method or a system to properly classify the documents in their respective domains. Different methods are used today along with manual classification of documents. Some methods are not useful when there are considerable number of documents to be classified. In this project, we plan to compare ontology-based text-mining (OTMM) approach with a method which uses feature extraction in hierarchical clustering based on ontology based on different performance parameters. Additionally, we propose a system which uses Hierarchical clustering method based on Ontology to cluster the documents in their respective domains. Using hierarchical clustering algorithms results in more accurate and efficient results. We intend to use text feature algorithm to improve the process of data retrieval process. This algorithm will help in removing the noises from the data set which will further help in improving the data retrieval process. The implemented system will support several types of documents like text files, pdf files, word document, images and also web files. Further the results will be visualised using visualization tools.

PAPER ID: DM17

Smart Health Prediction Malav Shah 1, Harsh Shah 2, Parth Oza 3, Dinesh Tharwani4




Whenever a person acquires some or the other symptom, he/she has to visit the doctor to determine the disease he/she has. Although there are sources like Google and other search engines, but none of them provides an answer for each and every related symptom instantly. The time taken by the patient to visit the doctor after acquiring the symptom can lead to escalation of a particular disease that they might be experiencing. The primary focus of our project is identification of diseases that are difficult to detect or diagnose at an early stage with the help of the patient’s medical history and the correlations among the symptoms that the patients might be experiencing. Our project intends to solve the existing discrepancies in the current system by


DJ ASCII – 18

35

improving the system even further. The current system does not provide the user to input particular symptoms and rather only gives him/her the broad outlook of the disease that they might be feeling based on some vague symptom. It only provides the diseases based on one or two of the symptoms, that too individually. It does not take the relation of various symptoms into account and this is the major issue being faced. This paper includes the research that will form the bedrock for the implementation phase of our project. Our motivations and objectives of creating a system that smartly predicts diseases based on related symptoms taking into account various parameters like Body Mass Index, lifestyle of an individual and prioritizing the type of disease the person may have been clearly defined in the report.

PAPER ID: DM19

Twitter Analysis of Trending Hashtags Utkarsh Dubey1, Varun Agarwal 2, Bhavya Shah 3

1,2,3, Department of Information Technology,


Twitter is extensively used for response to disaster management, both for information exchange and mapping the crisis, among citizens, and in relation to national and international humanitarian responders. This paper reports Twitter analysis aimed at identifying the most pressing issues that arose in the short term recovery phase starting about a week after the Nepal earthquake. Based on Twitter data collected between last week of April and first week of May 2015. 40,236 raw messages apparently related to the Nepal earthquake were retrieved, filtered and analyzed. This paper shows measures and analysis of the situation using geolocation feature for identifying danger zones and visual analytics of the dataset. The results show that our disaster module can work on different hashtags provided the data set is available for identify that unique problems.

DJ ASCII – 18

36

PAPER ID: DM22

Classifying Imbalanced Data Streams in the Presence of

Concept Drift Chandrasekhar Raman 1, Pratik Bhambhani2, Aqsa Bhimdiwala 3



E-mail: [email protected], [email protected], [email protected],

Classification is one of the most widely studied problems in the data mining and machine

learning communities. Applications involving streaming data are ubiquitous. Typical examples

include computer network traffic, phone conversations, ATM transactions, and soon. In data

streams, the properties of the dataset or the output variable tend to drift over time due to changes

in the environment and hence it becomes difficult to classify data in a non-stationary

environment. Therefore, we propose a solution for classifying data in a non-stationary

environment (data streams) where we have an imbalanced dataset and the underlying concept

that projects the attributes to the class labels is changing continuously viz concept drift. The

first layer is used to check whether the incoming data stream has class imbalanced tuples. This

information is passed on to the second layer which is tasked with detecting concept drift in the

stream. The procedure used for detection depends on whether imbalance was observed in the

previous layer. The knowledge generated from the second layer is passed onto the endmost stage

for the final level of computation. Here appropriate changes are made in the classification

algorithm such that the machine learning algorithm is not adversely affected by the drift.

PAPER ID: HCI03

3D Visualization and Colorization of Brain MR Images Dhruvesh Mehta1, Gitika Daswani2

1,2, Department of Computer Engineering,


E-mail: [email protected], [email protected]

There has been swift development of non-invasive brain imaging technologies which have led to various new beginnings in the analysis and study of brain anatomy and functioning. Magnetic resonance imaging (MRI) is largely responsible for the progress made so far in accessing brain injury and looking into its anatomy. Large amount of high quality data is available due to development in brain MR imaging. The examination of these huge and complex MRI datasets has turned into a tedious and complex errand for clinicians, who need to physically separate imperative data. This manual examination is frequently tedious and inclined to blunders because of different inter/intra-operator variability studies. Through this project, we propose


DJ ASCII – 18

37

computerized methods to improve disease diagnosis and testing. We intend to segment the grayscale MR images of the brain into distinct regions composed of pixels with the same characteristics based on prior anatomical knowledge of the three primary regions - grey matter, white matter and cerebrospinal fluid; colorize the resulting segments to enhance the perception and interpretation of the images and further process this data in order to generate a 3D visualization of the brain to facilitate the study of pathological regions for planning a surgery.

PAPER ID: HCI04

Attention Monitoring System in a Classroom Uma Sreeram 1, Aastha Joshij 2, Nikita Parmar 3



E-mail: 3 [email protected]

Educational institutions today try to impart quality education to its students however there is no efficient method to evaluate how well the lecture topics are grasped to the students. We propose to build an efficient method to assess the students’ attention level during the lecture to determine if they have understood the lecture or not. Our proposed system can be broken down into a series of steps: video to image frame generation, face detection, facial landmark detection, hand over face detection, classification and result analysis. By this the most prominent facial expressions and hand gestures and detected and used for classification as Paying Attention or Not Paying Attention. On the basis of this, the teachers can understand the students’ attention patterns and make modifications in their teaching methodology accordingly.

DJ ASCII – 18

38

PAPER ID: HCI08

Lung Cancer Detection Dhyanvi Jhaveri1, Manali Nagda 2

1,2, Dwarkadas J. Sanghvi College of Engineering, Mumbai, India

E-mail: 1 [email protected], 2 [email protected]

Lung cancer is a disease of abnormal cells multiplying and growing into a tumor. Lung Cancer is the leading cause of cancer deaths. So the process of early detection of the disease plays a very important and essential role to avoid serious advanced stages to reduce its percentage of distribution. The main aim is to detect the features and conclude whether the person is infected or not. . Hence, a lung cancer detection system is used to classify the presence of lung cancer in a CT-images. In this study, process such as image pre-processing, segmentation and feature extraction have been discussed in detail. We are aiming to get the more accurate results by using machine learning algorithms. Need of Automated System: To determine if there is a need for radiologists’ to continue to draw manual segmentations for these predictions, we propose an automated system and show that its predictive accuracy is comparable with the predictions obtained using an automated system based on the manual radiologists’ segmentations. The manual segmentations are time consuming and costly.

PAPER ID: HCI09

Oculus Vision for Blind Dennis Jayesh Mistry 1, Varun Vasudev Mukherjee2, Nayan Thakor Panchal3, Prof. Aruna U. Gawde4


Dwarkadas J. Sanghvi College of Engineering Mumbai, Maharashtra 400056


Oculus enables the user without vision to enable self-navigation without the requirement of the third person. The fundamental principle used is the concept of object detection by continuous capture of image by the Web-Camera. When there is scarcity of light the ultrasonic sensor will alert the person in case if they get too close to the obstacle. This paper demonstrates the basic configuration of the hardware units which can lead towards detecting the obstacles in the way and notifying the user about it so that necessary action can be taken forward to avoid the obstacle. The output is in the form of speech. The hardware used for implementing vision is by using the Raspberry pi 3 and a single Web-camera unit with ultrasonic sensor. The entire set up is mounted on a hat. In order to achieve this we make use of Open CV which is being implemented in raspbian OS. The aim is to enable low cost solution to the visually impaired and also enable them to move freely.

DJ ASCII – 18

39

PAPER ID: HCI10

Optic Interactive System Bhuvana Iyer 1, Krishnakant Thakur 2, Sanjana Kale 3, Shivani Jadhav 4


Ramrao Adik Institute of Technology, Nerul Dr. D.Y. Patil Vidyanagar, Sector-7, Nerul, Navi Mumbai


Human-Computer Interaction (HCI) have uncovered umpteen means for computer-user interaction. Using eye-gaze and eye movements as an input to a computer is a new evolving discipline of HCI which will add profound merit to this domain in future. Optic Interactive System is a low cost real- time pupil detection and tracking device which detects the position and movements of the pupil and moves the cursor on the computer screen accordingly. Currently, many commercial devices are developed for the same purpose but none of them are inexpensive or easily available. We aim to eradicate this problem of cost and availability by creating an inexpensive device using day-to-day equipments which will work similar to the commercial devices available in the market.

PAPER ID: HCI14

Smart Trainer: A Virtual Guide for Personal Training in

Exercise Twinkle Dhanak 1, Saheb Singh2, Niyati Maheshwari3




The current century has given rise to advancements in technology which has made the lifestyle of a number of people in the society sedentary. But with the availability of a number of video tutorials for fitness training and exercises available online, we can make our lifestyle healthier. To reap the benefits of these resources, a guide is needed to examine if these exercises are performed well. Also, not doing the exercises properly can have numerous side-effects on the human body whose results will be visible in the long run. A personal trainer is a person who does this job but not everyone can afford to have one. This paper suggests of a novel solution to the same problem using image processing and machine learning. The approach understanding the level of expertise of a user and the video of the user exercising suggest the necessary changes to the user in their posture. This will prevent side effects of attaining incorrect postures and at the same time maps and increases user benefits from the same. It has numerous applications in physiotherapy, self-training, and maintaining body-weight and a healthy lifestyle.

DJ ASCII – 18

40

PAPER ID: HCI15

VR HARAPPA Hem Acharya 1, Prutha Dubhashi 2, Rhea Sanghvi 3, Prof. Neha Katre 4



E-mail: 1 [email protected], 2 [email protected], 3 [email protected], [email protected]

We propose a system called VR HARAPPA as a VR based learning tool. It will enable students to learn history in a more immersive way. Due to the increased interest in augmented reality and Virtual Reality, there have been many applications developed. However, the current systems available in learning History are not immersive. It lacks participatory interaction from the user point of view. With VR HARAPPA, we aim to increase user interaction. Keeping this in mind, the primary goal of VR HARAPPA is to provide knowledge based on the syllabus of the ISCE and other education boards. Blender and Unity are open source 3d animation tools which can be used for the modelling and animations aspects. A VR headgear is used for creating a virtual environment. This interactive method can be used to make the whole concept of learning more visual and animated rather than the earlier systems. If the VR HARAPPA is implemented in any learning institute, the students will be able to learn in a uniquely and interactively by making the best use of the available technology.

PAPER ID: HCI17

An Educational Augmented Reality App to Enhance

Learning Experience Vineet Malhotra 1, Dhruvil Desai2, Hasan Banswarawala3

1,2,3 Department of Computer Engineering ,


E-mail: [email protected], 2 [email protected], 3

[email protected]

Augmented Reality is changing education in a dramatic way and it brings a new dimension to teaching and learning practices through visualization of the real world in an interactive environment. The aim of this research is focused at developing a mobile based Augmented Reality application using Vuforia and Unity which will be helpful and valuable for students in reinforcing their learning experience.

DJ ASCII – 18

41

PAPER ID: HCI18

VR Chem Lab Chintan Shah1, Sharang Ukidve2, Neal Gosalia3, Prof. Neha Katre4



E-mail: [email protected], 2 [email protected], 3 [email protected], [email protected]

This project aims at building an interactive environment where students can learn to perform chemical experiments. It addresses the overall problem of students especially college students lacking the understanding of chemical experiments. The environment will be a simulated virtual environment. The students will be able to interact with the objects in the environment, and perform experiments.

PAPER ID: NS01

Malware Generation using Taint Graph Ayushya Mishra 1, Mehul Sanghavi 2, Pawan Shah 3, Aman Gandhi 4



E-mail: [email protected], 2 [email protected], [email protected], [email protected]

Cyber security and information security acts as nucleus of any computer application

infrastructure. Cyber-space is constantly humbled owing to the vast array of vulnerability that

exist in any application residing in the system. Malware is an umbrella term for a lot of viruses

that are in the wild. Most attacks are widely targeted on capturing user’s (victim’s) sensitive

information and more or less can be detected by the available security measures like Antivirus,

Intrusion Detection Systems, etc. But when a sophisticated malware attack happens which

bypasses all the static signature based malware detection mentioned above it seeks a novella and

dynamic (read flexible) approach to be able to tackle with those attacks. These attacks may have

viruses which are polymorphic in nature or are zero-day attacks or they are simply multi-step

attacks. We propose a system capable of dynamically detecting malware that are key logger in

nature by generating signatures through taint graphs. Our system will log events in the

guest/victim system created by a running process and construct a log file from it which will have

comprehensive information of all action performed by that process in the system which is later

used to generate a taint graph or policy. This mentioned signature is intended to be flexible in

nature to also take in effect any minor or otherwise change in the attack since it relies on detecting

violation with the policies.



DJ ASCII – 18

42

PAPER ID: NS02

AWI Automatic Website Developer Mahavir Rathod 1, Aayushi Shah 2, Karan Vyas 3, Mitchell D’silva 4




In today’s world, developing website have become an integral part for the growth of any organization. A user planning for a startup wants to develop a website but cannot do so due to lack of technical knowledge. Even if the user hires a developer for developing the website; the content for the website which is the vital part for development of the website cannot be written properly by the user and hence he needs to hire someone wo can generate content for him. This is a laborious and an expensive process. There is some existing system such as mobilize, menu system, etc. to develop the website, but they don’t suffice all the needs of the user. Thus, this paper proposes a system which will try to overcome the above shortcomings by introducing AWI, which is a Chabot that will interact with the user and help him to build the website by selecting layouts and templates depending on the categories they choose. After which it will allow the user to edit the template, he has selected by interacting with the bot. It also gives the freedom to customize his website by selecting vivid headers, footers, images and section based on the category he has selected. Moreover, it also helps the user to automatically generate the content of the website by answering the questions depending on the type of category. The system will also perform search engine optimization for the user. Thus, it will save a lot of user’s time and money.

PAPER ID: NS05

Load Balancing in Software Defined Network Ashwini Swain 1, Karan Savla2, Kushal Ajmera 3



E-mail: [email protected], [email protected], 3 [email protected]

Software-defined networking (SDN) technology is an approach to computer networking that allows network administrators to programmatically initialize, control, change, and manage network behavior dynamically via open interfaces and abstraction of lower-level functionality. It separates data plane and control plane of network devices. The data plane represents all of the data that is being forwarded through the network like packets and the hardware that is used to forward it such as switches, routers. The control plane represents all logic and devices that are responsible for making forwarding decisions.



DJ ASCII – 18

43

PAPER ID: NS07

Data Encryption using Fibonacci Series and Unicode

Characters Prof. Prasad Tambekar 1, Prof. Vijay Shelke 2, Snigdha Mandal 3, Snehal Khachane4, Diksha Shetty 5


Yadavrao Tasgaonkar College of Engineering and Management, Chandai, Karjat


The objective of cryptography is to make it feasible for two persons to exchange a message in such a way that other persons cannot understand. There is no end to the number of ways this can be done, but here the proposed method will be more concerned with a technique of encoding the text in such a way that the recipient can only discover the original message. The original message usually called plain text is converted into cipher text by finding each character in the message and replacing it with another character based on the Fibonacci number generated. Further cipher text is converted into Unicode symbols, which avoid suspicion from the third party when send through an unsecured communication channel. There are two levels in the proposed method; (i) converting plain text to cipher text and (ii) converting cipher text to Unicode symbols. In each level, security key is used to encode the original message which provides two levels of security from intruders. On the other end, the extraction algorithm is designed in such a way that the process converts the Unicode symbols into cipher text and then cipher text to plain text. This encoding and decoding scheme of the proposed method is significantly different as compared to the traditional methods.

PAPER ID: NS08

No Cost Real-time Content Management System Emmanuel Francis Kolengaden1

[email protected]

The matter of basic CRUD (Create Read Update Delete) operations are a major necessity for any

organization, be it Corporate or a Small Business. But high speed data transmission and

synchronization of data comes at a cost which only corporates tend to enjoy. Well, my Real-time

system comes to your aid at times where data can be accessed from any app (iOS , Android ,

Windows) or Website . Data stays synchronized and that too with 128 Bit SSL that makes every

bit of data encrypted at every stage. This is possible when multiple Opensource (IONIC 3,

ANGULAR 5 (By Google), NODEJS, Firebase (A High Speed RESTful Service by Google)

platforms come together to form the heart of this system. The implemented system has the ability

to instantly manage content on your App or Website in a matter of milliseconds unlike PHP and

.NET (which takes 3-5 seconds) that requires several lines of code to perform the same operation

DJ ASCII – 18

44

with the help of 3rd party APIs . In order to make changes to the Database only one line of code

needs to be changed. The ease of operation and the time taken for the transactions are blazing fast

as all the APIs used in this project are hosted on Firebase. The Realtime CMS is 99% OpenSource

, Highly Scalable , Object Oriented and data is instantly updated across multiple platforms and

devices without the need for any PAID services out there . And it doesn’t get any better when

you hear that the my CMS is a cross platform HYBRID App, hence the same code for iOS ,

Android and Windows platforms. My system helps you spend more time on the core

functionality of the App than the UI Aspects. So apps like Microblogs, Multiplatform Chat App ,

Registration Management System , App Remote Control (Remotely Update Data on App) and

Web Controller ( Remotely Update Content on Websites) in Real-time could be easily built within

a fortnight with my implemented system. Basically , every element within the my No Cost Real-

time CMS can be easily scaled and synchronized as per the client’s needs , without the need for

painstaking JAVA , PHP , SPRING or .NET . Many Companies like CBS, Citrix, Koding ( Realtime

Cloud VMs) are switching to services similar to my system due to its mere simplicity , scalability

and support.

PAPER ID: NS16

Lightweight Authentication and Encryption Mechanism in

Routing Protocol for Low Power and Lossy Networks Akshay Shah 1, Nishit Sakariya 2, Dishith Poojary 3



Newer concepts like Wireless Sensor Networks (WSN), Internet of Things (IoT), 6LoWPAN (IPv6 over Low-Power Wireless Personal Area Networks) etc are emerging. So, protocols, techniques and concepts that have been developed for the traditional computational devices do not essentially apply to the newer low power, less capable devices which are an integral part of the Internet of Things paradigm. So newer protocols need to be developed and [1] has drafted in specifications for the routing protocol for Low Power and Lossy Networks. LLN routers typically operate with constraints on processing power, memory, and energy (battery power). Their interconnects are characterized by high loss rates, low data rates, and instability. We aim to propose an authentication and encryption protocol for these Low Power and Lossy Networks (LLN). We have chosen PRESENT as the encryption algorithm and SQUASH as the Authentication Algorithm.

DJ ASCII -18

Documents

Transcript of DJ ASCII -18