By Xuanyi Dong, Deyu Meng, Fan Ma, Yi Yang. This paper is accepted by ACM Multimedia 2017.
Dual-Network is a weakly supervised object detection framework leveraging deep CNN models.
This project is modified on the Matlab code of R-FCN and Fast R-CNN.
Dual-Network is released under the MIT License (refer to the LICENSE file for details).
- ImageNet-pretrained networks: Google Drive. Please save the models into the corresponding sub-directory of
models/pre_trained_models
. - The initial pseudo labels for PASCAL VOC 2007 by ContextLocNet : Google Drive. Please save and extract it into
data
. - The pre-computed region proposals: Google Drive. Please save and extract it into
data
. - Download the PASCAL VOC 2007 data into
datasets
, following the README indatasets
. - Compile
Caffe
located inexternal/caffe
. - Run
dual_build.m
to complie the nms mex functions. - Run
startup.m
to add necessary paths.
- [TODO] re-organize the experiment codes.
If you find Dual-Network useful in your research, please consider citing:
@inproceedings{dong2017dual,
title={A Dual-Network Progressive Approach to Weakly Supervised Object Detection},
author={Dong, Xuanyi and Meng, Deyu and Ma, Fan and Yang, Yi},
booktitle={Proceedings of the 2017 ACM on Multimedia Conference},
pages={279--287},
year={2017},
organization={ACM}
}
@inproceedings{kantorov2016,
title = {ContextLocNet: Context-aware Deep Network Models for Weakly Supervised Localization},
author = {Kantorov, V., Oquab, M., Cho M. and Laptev, I.},
booktitle = {Proc. European Conference on Computer Vision (ECCV), 2016},
year = {2016}
}
@article{dai16rfcn,
Author = {Jifeng Dai, Yi Li, Kaiming He, Jian Sun},
Title = {{R-FCN}: Object Detection via Region-based Fully Convolutional Networks},
Journal = {arXiv preprint arXiv:1605.06409},
Year = {2016}
}
@inproceedings{girshick2015fast,
title={Fast R-CNN},
author={Girshick, Ross},
booktitle={Proceedings of the IEEE international conference on computer vision},
pages={1440--1448},
year={2015}
}