DANet: Divergent Activation for Weakly Supervised Object Localization，in ICCV 2019

We propose a divergent activation (DA) approach, and target at learning complementary and discriminative visual patterns for image classification and weakly supervised object localization from the perspective of discrepancy. To this end, we design hierarchical divergent activation (HDA), which leverages the semantic discrepancy to spread feature activation, implicitly. We also propose discrepant divergent activation (DDA), which pursues object extent by learning mutually exclusive visual patterns, explicitly.

1. Clone this repo:

   DANet_ROOT=/path/to/clone/DANet
   git clone --recursive https://github.com/xuehaolan/DANet $DANet_ROOT
   cd $DANet_ROOT
   

2. Create an Anaconda environment with python2.7 and PyTorch>=0.4.0

Download the images of CUB-200-2011 dataset and place the data at $DANet_ROOT/data/CUB-200-2011

• Train

  cd scripts
  sh train_DA_cub.sh
  

• Test

  Download the pretrained model at GoogleDrive(https://drive.google.com/open?id=114zDDJe6zYbRjashntibbN1D4DM0E1uw)

  cd scripts
  sh val_DA_cub.sh
  

• Note

1. Using modified VGG(vgg_DA_p) could achieve higher performance for both CAM and DANet.
2. The valiation code uses simple thresholding, using localization method provided by CAM may gets greater localization results.

In this project, we reimplemented CHR on PyTorch based on SPG.

Please consider citing our paper in your publications if the project helps your research.

@inproceedings{xue2019danet,
 title={Danet: Divergent activation for weakly supervised object localization},
 author={Xue, Haolan and Liu, Chang and Wan, Fang and Jiao, Jianbin and Ji, Xiangyang and Ye, Qixiang},
 booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
 pages={6589--6598},
 year={2019}
}