@inproceedings{97a0aa4bf86d4fad955743e457cb0653,
title = "Object recognition based on efficient sub-window search",
abstract = "We propose a new method for object recognition in natural images. This method integrates bag of features model with efficient sub-window search technology. sPACT is introduces as local feature descriptor for recognition task. It can capture both local structures and global structures of an image patch efficiently by histogram of Census Transform. An efficient sub-window search method is adapted to perform localization. This method relies on a branch-and-bound scheme to find the global optimum of the quality function over all possible sub-windows. It requires much fewer classifier evaluations than the usually way does. The evaluation on PASCAL 2007 VOC dataset shows that this object recognition method has many advantages. It uses weakly supervised training method, yet has comparable localization performance to state-of-the-art algorithms. The feature descriptor can efficiently encode image patches, and localization method is fast without losing precision.",
keywords = "Bag of features, Object detection, Object recognition, sPACT",
author = "Qing Nie and Shouyi Zhan and Weiming Li",
year = "2009",
doi = "10.1007/978-3-642-05253-8_48",
language = "English",
isbn = "3642052525",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "435--443",
booktitle = "Artificial Intelligence and Computational Intelligence - International Conference, AICI 2009, Proceedings",
note = "International Conference on Artificial Intelligence and Computational Intelligence, AICI 2009 ; Conference date: 07-11-2009 Through 08-11-2009",
}