Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf ·...
Transcript of Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf ·...
![Page 1: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/1.jpg)
Object DetectionThrough Machine Learning
Machine Learning and Applications Group, 2018.
Uroš Stegić
![Page 2: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/2.jpg)
TRADITIONAL COMPUTER VISION
General Overview
Convolution Operator
Filters
Convolutions Over Volume
![Page 3: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/3.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Description
Process & analyze visual signalExtract information from visual signalPerform on raw signal (pixel intensities values)
3
![Page 4: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/4.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Enhance Intuition
Computer Graphics
(x0, y0) = (3, 4)r = 6↓
Computer Vision
↓(x0, y0) = (3, 4)
r = 6
4
![Page 5: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/5.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Tasks in Computer Vision
Object RecognitionImage RetrievalObject DetectionOCRPose Estimation...
TrackingScene ReconstructionOptical FlowSemantic SegmentationImage Reconstuction...
5
![Page 6: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/6.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Tasks in Computer Vision
Object RecognitionImage RetrievalObject DetectionOCRPose Estimation...
TrackingScene ReconstructionOptical FlowSemantic SegmentationImage Reconstuction...
6
![Page 7: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/7.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Convolution Operator - Definition
DefinitionLet A,B ∈ D ⊆ Rn×n. Convolution operator, denoted as ∗maps the spaceD×D to a field of real numbers and is definedas follows:
A ∗ B =
n∑i=1
n∑j=1
AijBij
7
![Page 8: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/8.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Convolution Operator - Example
1 2 34 5 67 8 9
∗
0 1 01 0 10 1 0
8
![Page 9: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/9.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Convolution Operator - Example
1 2 34 5 67 8 9
∗
0 1 01 0 10 1 0
= 2 ∗ 1 + 4 ∗ 1 + 6 ∗ 1 + 8 ∗ 1 = 20
9
![Page 10: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/10.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Convolution Operator - Example
1 2 34 5 67 8 9
∗
0 1 01 0 10 1 0
= 2 ∗ 1 + 4 ∗ 1 + 6 ∗ 1 + 8 ∗ 1 = 20
10
![Page 11: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/11.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Convolution Operator - Example
1 2 34 5 67 8 9
∗
0 1 01 0 10 1 0
= 2 ∗ 1 + 4 ∗ 1 + 6 ∗ 1 + 8 ∗ 1 = 20
11
![Page 12: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/12.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Convolution Operator - Example
1 2 34 5 67 8 9
∗
0 1 01 0 10 1 0
= 2 ∗ 1 + 4 ∗ 1 + 6 ∗ 1 + 8 ∗ 1 = 20
12
![Page 13: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/13.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Convolution Operator - Example
1 2 34 5 67 8 9
∗
0 1 01 0 10 1 0
= 2 ∗ 1 + 4 ∗ 1 + 6 ∗ 1 + 8 ∗ 1 = 20
13
![Page 14: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/14.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Filters
211 39 200 102 174 25 90 144138 44 184 110 193 30 92 136151 73 190 114 189 41 105 128129 101 123 181 201 169 117 191140 122 153 231 209 157 124 113221 115 77 244 198 149 156 247
∗
0 1 01 0 10 1 0
14
![Page 15: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/15.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Filters - Examples
Vertical Edge ExtractorHorizontal Edge ExtractorSobel filterSharpenGaussian Blur
15
![Page 16: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/16.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Filters - Edge Extractor
∗
1 0 −11 0 −11 0 −1
=
16
![Page 17: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/17.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Filters - Edge Extractor
∗
1 0 −11 0 −11 0 −1
=
17
![Page 18: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/18.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Filters - Sobel
∗
1 0 −12 0 −21 0 −1
=
18
![Page 19: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/19.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Filters - Gaussian Blur
∗ 1
256
1 4 6 4 14 16 24 16 46 24 36 24 64 16 24 16 41 4 6 4 1
=
19
![Page 20: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/20.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Multiple Input Channels
Figure: Convolution of multichannel image
20
![Page 21: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/21.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Multiple Filters
Figure: Convolution of multichannel image with two filters21
![Page 22: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/22.jpg)
CONVOLUTIONAL NEURALNETWORKS
Parameter Learning
Basic CNNs
Residual Networks
Inception Networks
![Page 23: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/23.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Basic Concepts
Figure: Convolutional layers stacked
23
![Page 24: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/24.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Basic Concepts - Takeaway
Image ClassificationParameters (filters) Learning [LBD+89]Weight SharingFeature Extraction
24
![Page 25: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/25.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Basic Concepts - Feature Abstractions
Figure: Feature Visualization [ZF13]
25
![Page 26: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/26.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Basic Concepts - Pooling Layers
Sampling important FeaturesReduce Computation TimeMake Features Robust
26
![Page 27: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/27.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Basic Concepts - Pooling Layers (Example)
Pooling Layer - Max Pooling
9 2 4 13 1 8 24 5 9 25 6 0 1
−→[9 86 9
]
27
![Page 28: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/28.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Basic Concepts - Architecture
Figure: Convolutional Neural Network - Example
28
![Page 29: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/29.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
CNN Architecture - Lenet-5
Figure: Lenet-5 Architecture [LBBH98]
29
![Page 30: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/30.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
CNN Architecture - VGG
Figure: VGG Architecture
30
![Page 31: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/31.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
CNN Architecture - AlexNet
Figure: AlexNet Architecture [KSH12]
31
![Page 32: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/32.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
CNN - Problems
Vanishing GradientExploding GradientComputational Complexity
32
![Page 33: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/33.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Residual Block
Figure: Residual Block (Skip Connection) [HZRS15]
33
![Page 34: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/34.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Residual Network
Figure: CNN Architecture - ResNet-34 [HZRS15]
34
![Page 35: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/35.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
1x1 Convolution
Figure: 1x1 Convolution [LCY13]
35
![Page 36: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/36.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Inception Module - Idea
Figure: Inception Module Naive Version [SLJ+14]
36
![Page 37: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/37.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Inception Module - Redone
Figure: Inception Module With Dimension Reduction [SLJ+14]
37
![Page 38: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/38.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Inception Network
Figure: Inception Network (GoogLeNet) [SLJ+14]
38
![Page 39: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/39.jpg)
OBJECT DETECTION
Task Outline
YOLO
RCNN Family
Other Influental Models
Speed/Accuracy Trade-Off
![Page 40: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/40.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Visualizing the Task
Figure: Object Detection Task40
![Page 41: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/41.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Understanding the Bounding Box Error
Figure: Bounding Box Missmatch
41
![Page 42: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/42.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Defining the IoU
Figure: Intersection over Union
42
![Page 43: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/43.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Gaining Intuition on IoU
Figure: Intersection over Union - Example
43
![Page 44: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/44.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Similar Bounding Boxes Problem
Figure: Elimination of Multiple Bounding Boxes
44
![Page 45: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/45.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Non-Maximum Suppression
Threshold every bounding boxSort bounding boxes by detection probability indecresing orderFor each bounding box bi remove all bounding boxesbj(j = i) such that IoU(bi,bj) ≥ t for some fixed t
45
![Page 46: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/46.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
YOLO - Introduction
Figure: Grid for YOLO
y =
pcbxbybwbhc1c2...cn
0You Only Look Once: Unified, Real-Time Object Detection [RDGF15]
46
![Page 47: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/47.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Limitations (already?)
Problem: Multiple objects centered in same cell
47
![Page 48: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/48.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Anchor Boxes
Choose a number of anchors (predefined bboxes)Select a ratio (width and height) for each of themModify the output to include this anchors...Profit
48
![Page 49: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/49.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Anchor Boxes - Example
y1 =
pc1bx1by1bw1bh1c11...cn1
, y2 =
pc2bx2by2bw2bh2c12...cn2
, y =
[y1y2
]
49
![Page 50: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/50.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
YOLO - Loss Fucntion
L(y, y) = λcoord
s2∑i=0
B∑j=0
1objij [(xi − xi)2 + (yi − yi)2]
+ λcoords2∑i=0
B∑j=0
1objij [(
√wi −√
wi)2 + (
√hi −
√hi)
2]
+s2∑i=0
B∑j=0
1objij (Ci − Ci)
2
+ λnoobjs2∑i=0
B∑j=0
1objij (Ci − Ci)
2
+s2∑i=0
1obji
∑c∈classes
(pi(c)− pi(c))2
50
![Page 51: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/51.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Region Based Approach
Propose Regions of InterestClassify each RoIRegress Bounding Box Coordinates
51
![Page 52: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/52.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Region Models
Regions with CNN (R-CNN) [GDDM13]Fast R-CNN [Gir15]Faster R-CNN [RHGS15]Mask R-CNN [HGDG17]
52
![Page 53: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/53.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Region Proposals - Selective Search
Figure: Selective Search Algorithm Visualized
53
![Page 54: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/54.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
R-CNN
Figure: R-CNN Pipeline
54
![Page 55: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/55.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Fast R-CNN
Convolution Based Sliding WindowROI PoolingSoftmax Classification
55
![Page 56: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/56.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Fast R-CNN - Sliding Window
Figure: Sliding Window - CNN Implementation
56
![Page 57: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/57.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Fast R-CNN - Visualized
Figure: Fast R-CNN Pipeline
57
![Page 58: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/58.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Fast R-CNN - Loss
L(p,u, tu, v) = Lcls(p,u) + λ[u ≥ 1]Lloc(tu, v)
Lcls(p,u) = − logpu
Lloc(tu, v) =∑
i∈{x,y,w,h}smoothL1(tui − vi)
smoothL1(x) ={0.5x2, if x ≤ 1
x− 0.5, otherwise
58
![Page 59: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/59.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Faster R-CNN
Bottleneck: Region Proposals by Selective Search (2s)Solution: Region Proposals by CNN (0.01s)
59
![Page 60: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/60.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Region Proposal Network
Figure: Region Proposal Network for Faster R-CNN60
![Page 61: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/61.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
RPN - Loss
L(pi, ti) =1
Ncls
∑iLcls(pi,p∗i ) + λ
1
Nreg
∑ip∗i Lreg(ti, t∗i )
61
![Page 62: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/62.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Faster R-CNN - Architecture
Figure: Model Scheme of Faster R-CNN
62
![Page 63: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/63.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Mask R-CNN
Figure: Model Scheme of Faster R-CNN
63
![Page 64: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/64.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Other Influential Models
RetinaNet (Focal Loss) [LGG+17]Single Shot Detector [LAE+15]
64
![Page 65: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/65.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
RetinaNet
Figure: Retina Net - Overview
65
![Page 66: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/66.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Speed vs. Precision
Figure: GPU Time vs. Precision [HRS+16]
66
![Page 67: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/67.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
Lecture Pronouncement
CONVERGENCE
67
![Page 68: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/68.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
References I
Ross B. Girshick, Jeff Donahue, Trevor Darrell, andJitendra Malik, Rich feature hierarchies for accurate objectdetection and semantic segmentation, CoRRabs/1311.2524 (2013).Ross B. Girshick, Fast R-CNN, CoRR abs/1504.08083(2015).Kaiming He, Georgia Gkioxari, Piotr Dollár, and Ross B.Girshick, Mask R-CNN, CoRR abs/1703.06870 (2017).
68
![Page 69: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/69.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
References II
Jonathan Huang, Vivek Rathod, Chen Sun, MenglongZhu, Anoop Korattikara, Alireza Fathi, Ian Fischer,Zbigniew Wojna, Yang Song, Sergio Guadarrama, andKevin Murphy, Speed/accuracy trade-offs for modernconvolutional object detectors, CoRR abs/1611.10012(2016).Kaiming He, Xiangyu Zhang, Shaoqing Ren, and JianSun, Deep residual learning for image recognition, CoRRabs/1512.03385 (2015).
69
![Page 70: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/70.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
References III
Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton,Imagenet classification with deep convolutional neuralnetworks, Advances in Neural Information ProcessingSystems 25 (F. Pereira, C. J. C. Burges, L. Bottou, andK. Q. Weinberger, eds.), Curran Associates, Inc., 2012,pp. 1097–1105.
Wei Liu, Dragomir Anguelov, Dumitru Erhan, ChristianSzegedy, Scott E. Reed, Cheng-Yang Fu, andAlexander C. Berg, SSD: single shot multibox detector,CoRR abs/1512.02325 (2015).Y. Lecun, L. Bottou, Y. Bengio, and P. Haffner,Gradient-based learning applied to document recognition,IEEE (1998), 2278–2324.
70
![Page 71: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/71.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
References IV
Yann Lecun, Bernhard Boser, John Denker, DonHenderson, R E. Howard, W.E. Hubbard, and LarryJackel, Backpropagation applied to handwritten zip coderecognition, Neural Computation 1 (1989), 541–551.
Min Lin, Qiang Chen, and Shuicheng Yan, Network innetwork, CoRR abs/1312.4400 (2013).Tsung-Yi Lin, Priya Goyal, Ross B. Girshick, Kaiming He,and Piotr Dollár, Focal loss for dense object detection,CoRR abs/1708.02002 (2017).Joseph Redmon, Santosh Kumar Divvala, Ross B.Girshick, and Ali Farhadi, You only look once: Unified,real-time object detection, CoRR abs/1506.02640 (2015).
71
![Page 72: Object Detection - Through Machine Learningmachinelearning.math.rs/Stegic-ObjectDetection.pdf · 2018-11-28 · TraditionalComputerVision ConvolutionalNeuralNetworks ObjectDetection](https://reader035.fdocuments.us/reader035/viewer/2022081522/5f2a4650ba291c0d7a646d4d/html5/thumbnails/72.jpg)
Traditional Computer Vision Convolutional Neural Networks Object Detection
References V
Shaoqing Ren, Kaiming He, Ross B. Girshick, and JianSun, Faster R-CNN: towards real-time object detection withregion proposal networks, CoRR abs/1506.01497 (2015).
Christian Szegedy, Wei Liu, Yangqing Jia, PierreSermanet, Scott E. Reed, Dragomir Anguelov, DumitruErhan, Vincent Vanhoucke, and Andrew Rabinovich,Going deeper with convolutions, CoRR abs/1409.4842(2014).Matthew D. Zeiler and Rob Fergus, Visualizing andunderstanding convolutional networks, CoRRabs/1311.2901 (2013).
72