CA_Code.tar.gz_CA matlab_color_the code

This folder contains the Matlab code for the Color Attention algorithm proposed in ICCV'09 paper "Top-Down Color Attention for Object recognition" [1]. The code is only provided as a guideline for implementation. As an example we provide the results on the SOCCER data set [3]. The algorithm first computes the probability of all the categories provided the color visual words. These probabilities are then further used to compute a weighted histogram by sampling the shape visual words. Input: The algorithm requires: a) The color visual word indexes based on Hue descriptor [2]. b) The Shape visual word indexes based on SIFT descriptor. The indexes are stored in a cell structure where each element refers to one image in the data set. These have been precomputed. First feature locations and scales have been detected with an interest point detector. Next for all features in the image the indexes have been computed by assigning the descriptors to color and shape visual words. Output: The Color Attention histogram. Run Example_CA for an example on how to use the algorithm. For any questions, contact me: Fahad Shahbaz Khan email: fahad@cvc.uab.cat [1] F. Shahbaz Khan, J. van de Weijer, M. Vanrell, Top-Down Color Attention for Object Recognition, Proc. ICCV, Kyoto, Japan, 2009. [2] J. van de Weijer, C. Schmid, Coloring Local Feature Extraction , Proc. ECCV, Part II, 334-348, Graz, Austria, 2006. [3] http://lear.inrialpes.fr/people/vandeweijer/data