Class24 computer vision 计算机视觉.pdf

EECE 5639 Computer Vision I Lecture 23 BoW: Implicit Shape Model Deep Learning Hw 5 is out. Now Due April 19 Next Class Final Review 1 Object Bag of ‘words’ 2 learning recognition feature detection codewords dictionary representation image representation category models category (and/or) classifiers decision 3 Problem with bag-of-words All have equal probability for bag-of-words methods Location information is important 4 Bag of Words and Spatial Information A bag of words throws away spatial relationships between features. Middle ground: Visual “phrases”: frequently co-occurring words Semi-local features: describe configuration, neighborhood Let position be part of each feature Count bags of words only within sub-grids of an image After matching, verify spatial consistency 5 Parts Structure 6 Implicit Shape Model Use Hough Space Voting to find object: Learn spatial distributions of the words wrt to a “reference point” Use HT to vote for reference points in the test image 7 Implicit shape models • Visual vocabulary is used to index votes for object position • [a visual word = “part”] visual codeword with
 displacement vectors training image annotated with object localization info B. Leibe, A. Leonardis, and B. Schiele, Combin