CAM based on ResNet50

0. Introduction :

• CAM on CUB-200-2011 dataset
• CAM backbone : ResNet50 (pretrained on ImageNet)

1. Virtual Env

conda :

 $ bash install.sh 

2. Preprocessing CUB-200-2011 dataset

 $ cd data/CUB_200_2011
 $ wget https://data.caltech.edu/records/65de6-vp158/files/CUB_200_2011.tgz
 $ tar -xzvf CUB_200_2011.tgz
 $ python preprocessing.py

3. Training

 $ bash exp.sh 

4. Evaluation

 $ bash eval.sh

5. Visualization for one image

 $ python visualized_cam.py --weights work_dir/cub/vgg16_cam_0.001:45 --image-id 1

6. References :

• Zhou, B., Khosla, A., Lapedriza, A., Oliva, A., Torralba, A.: Learning deep features for discriminative localization.
• https://github.com/jiwoon-ahn/irn
• https://github.com/zhoubolei/CAM/tree/c63f2850a7a3dadc21fa1b021875e2d4d053ece5
• https://github.com/zhangyongshun/resnet_finetune_cub
• https://github.com/Uason-Chen/CTR-GCN
• https://github.com/won-bae/rethinkingCAM
• https://github.com/junsukchoe/ADL