Part-based RCNNs for Fine-grained Category Detection

This work is created by Ning Zhang, Jeff Donahue, Ross Girshick and Trevor Darrell from UC Berkeley.

Citing this work

If you are using this code for your research, please cite the following paper:

@inproceedings{ZhangECCV14,
    Author = {Zhang, Ning and Donahue, Jeff and Girshick, Ross and Darrell, Trevor},
    Title = {Part-based RCNN for Fine-grained Detection},
    Booktitle = {European Conference on Computer Vision},
    Year = {2014}
}

License

This software is under BSD 3-Clause License, please refer to LICENSE file.

Prerequisites

0. Caffe

• Download caffe from http://caffe.berkeleyvision.org/ and follow the instructions to install.
• Change caffe matlab wrapper path in init.m

0. RCNN

• Download source code from https://github.com/rbgirshick/rcnn and follow the instructions to install.
• Change rcnn path in init.m
• In order to train part detectors for CUB2011 dataset, replace the following three functions in imdb/imdb_from_voc.m imdb/roidb_from_voc.m and imdb/imdb_eval_voc.m to the functions in imdb_cub folder.
• Follow rcnn instructions to train the part detectors.

0. Liblinear

• Download liblinear package from http://www.csie.ntu.edu.tw/~cjlin/liblinear/

Annotation/ has annotated part boxes on CUB200-2011 dataset.

Usage

• run.m is the main function to reproduce the results in the paper.
• Part detectors, finetuned models, feature representations are cached. Download the cache files by running get_cache_files.sh and unzip to caches/ folder.

###Bug report If you have any issues running the codes, please contact Ning Zhang (nzhang@eecs.berkeley.edu).