HOGgles Visualizing Object Detection Features C Vondrick A Khosla T Malisiewicz A Torralba ICCV 2013 presented by Ezgi Mercan Object Detection Failures Why do our detectors think water looks like a car ID: 762242
Download Presentation The PPT/PDF document "HOGgles Visualizing Object Detection Fe..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
HOGglesVisualizing Object Detection Features C. Vondrick , A. Khosla , T. Malisiewicz , A. Torralba ICCV , 2013 . presented by Ezgi Mercan
Object Detection Failures Why do our detectors think water looks like a car? D ata set?Machine learning method?Features? 10/11/2013 CSE590V 2 A high scoring car detection from DPM.
Object Detection Failures 10/11/2013 CSE590V 3 Object Categories Aeroplane Table Bicycle Dog Bird Horse Boat Motorbike Bottle Person Bus Potted plantCarSheepCatSofaChairTrainCowTV/ monitor Person Chair Car
Reconstructing an Image 10/11/2013 CSE590V 4 original image copied patches blending interpolation Reconstructing an image from its local descriptors, Philippe Weinzaepfel , Hervé Jégou and Patrick Pérez, Proc. IEEE CVPR’11. calculate SIFT elliptic region of interestaffine normalization to square patchlocal descriptor
Histogram of Oriented Gradients 10/11/2013 CSE590V 5 Histograms of oriented gradients for human detection, Dalal, N . , Triggs , B., CVPR’05. original image R-HOG descriptor (HOG glyph) R-HOG descriptor weighted by positive SVM weights R-HOG descriptor weighted by negative SVM weights
Algorithms10/11/2013 CSE590V 6 Image: HOG descriptor: HOG inverse: Exemplar LDA Ridge Regression
Algorithms Direct Optimization Image: HOG descriptor: Image basis: Coefficients: 10/11/2013 CSE590V 7
Algorithms Paired Dictionary Learning Image: HOG descriptor: Image basis: HOG basis: Coefficients: 10/11/2013 CSE590V 8
Paired Dictionary Learning 10/11/2013 CSE590V 9 HOG feature HOG Basis Image Basis HOG Inversion
Paired Dictionary Learning10/11/2013 CSE590V 10 Solving paired dictionary learning problem: s.t. , some learned pairs of dictionaries for and
Feature Visualization10/11/2013 CSE590V 11 Origina l ELDA Ridge Regression Direct Optimization Paired Dict. Learning
10/11/2013 CSE590V 12
10/11/2013 CSE590V 13
Deformable Parts Model 10/11/2013 CSE590V 14 Potted Plant Chair Does HOG capture color? Object detection with discriminatively trained part-based models, P . Felzenszwalb , R. Girshick , D. McAllester , and D. Ramanan . PAMI, 2010.
Computers can see better than us10/11/2013 CSE590V 15
Questions?10/11/2013 CSE590V 16 Visit web.mit.edu/ vondrick /ihog / for more cool stuff.