Title: Fast Concurrent Object Localization and Recognition
Author: Tom Yeh, John J. Lee, Trevor Darrell.
Year: CVPR 2009
vj(i) : the feature j 's contribution for i-th object(i-th object model)
pull down all visual words in all images into a quantization clustering tree.
So, each visual word can get its contribution score by looking up the histogram(distribution)
of the clustering tree. The histogram's dimension is M , which M is how many objects(books,butterflies..etc) in this dataset. (different M books..etc)
ex: 1 5 7 9 2 0
b1 b2 b3 b4 b5 b6
For a query image, each visual word of
ISM: implicit shape model
working ...
0 Comment(s):
Post a Comment