Object Detection Using the Statistics of Parts
-
Upload
orson-case -
Category
Documents
-
view
15 -
download
2
description
Transcript of Object Detection Using the Statistics of Parts
![Page 1: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/1.jpg)
Object Detection Using the Statistics of Parts
Henry Schneiderman
Takeo Kanade
Presented by : Sameer Shirdhonkar
December 11, 2003
![Page 2: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/2.jpg)
Overview Main Features of Paper
• Multiple Exhaustive Classifiers
• Parts based representation :Discretized Wavelet Coefficients
• Estimating probabilities :AdaBoost with Confidence Weighted Predictions
![Page 3: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/3.jpg)
Classifier Design
• Part : Set of input features which are statistically inter-dependent, and independent of other features.
• Wavelet Coefficients as Features: Linear Phase 5/3 perfect reconstruction filter bank– Invertible transform [ but not after quantization ]– Partially decorrelates natural scenes – less features
needed– Parts can be localized by space, frequency and
orientation– Multiresolution nature speeds up computation
![Page 4: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/4.jpg)
Classifier Form
• Likelihood Ratio Test [ Used similar to SPRT ]
• Generalization of Ideal Classifier Table[ Object present/absent for all possible feature values ]
• Convert P(Image|Object) and P(Image|Non-Object) to P(object|mage)
• Change P(Object|Image) to Classifier output (present/absent)
![Page 5: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/5.jpg)
Approximations
• Parts are statistically Independent – Localized Dependence for cars, faces, etc.
• Part values (Wavelet Transform coefficients) are quantized
• Part positions are quantized coarsely
![Page 6: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/6.jpg)
![Page 7: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/7.jpg)
Local Operators• Locality in position more important
• Local Operator – Moving Combination of Wavelet coefficients
![Page 8: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/8.jpg)
Local Operator Design
• Intra-subband operators – 6– Joint localization in space, frequency and orientation
• Inter-Orientation operators – 4– Localization in space and frequency, different orientations
• Inter-frequency operators – 6– Localization in space and orientation, broad frequency
content
• Inter-Orientation + Inter-Frequency Operator – 1 – Localization in space, different frequency and orientation
![Page 9: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/9.jpg)
The Hard Part: Collecting Data
• Pre-processing Object Images:– Size normalization and Spatial Alignment– Intensity Normalization and Lighting
Correction – Separate normalizations for left and right parts of face (5 discrete values)
– Synthesizing data : Positional perturbation, Overcomplete evaluation of wavelet transform, background substitution, low pass filtering
• Non-object images : Bootstrapping
![Page 10: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/10.jpg)
Training
• Probabilistic Approximation– Filling the histogram bins of Parts
• AdaBoost :– Train Multiple Classifiers ht(x) with weighted training samples.– First Classifier h1(x) – equal weights to all.– Next – Higher weight to Incorrectly classified samples
– Final Classifier:
– αt found by binary search– The weighted sum of classifiers is reduced to a single classifier due to
linearity (in log likelihood).– Use Cross Validation to prevent overfitting
![Page 11: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/11.jpg)
Efficient Exhaustive Search [Does this exist ?]
• Algorithm uses exhaustive search across position, size, orientation, alignment and intensity.
• Course to Fine Evaluation – similar to SPRT• Wavelet Transform coefficients can be reused for
multiple scales• Color preprocessing• Time – 5 s for 240x256 image (PII 450 MHz)
![Page 12: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/12.jpg)
Results : Face DetectionSometimes it Works
![Page 13: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/13.jpg)
And Sometimes it Doesn’t
![Page 14: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/14.jpg)
Results : Car Detection
![Page 15: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/15.jpg)
DiscussionWhich are the Important Parts ?
![Page 16: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/16.jpg)
Conclusion
• Works pretty well
• Training is difficult and needs too much manual intervention
• Slow – due to exhaustive search
![Page 17: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/17.jpg)
How many faces in this picture ?
![Page 18: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/18.jpg)
![Page 19: Object Detection Using the Statistics of Parts](https://reader035.fdocuments.us/reader035/viewer/2022070400/5681354d550346895d9ca8ee/html5/thumbnails/19.jpg)
What about this ?