Codeczh / OPAM_python

OPAM_rerun

Dataset:

If you want to use CAR196 dataset, download CAR196/images from here: CAR196, and put /images in CAR196 folder.

While if you want CUB200 dataset, download CUB200/images from here: CUB200, or use CUB200/split_train_test.py to process your original CUB_200_2011 and generate these images.

Note: If you want your own dataset, remember to edit your config.yaml and change classnum: xxx.

---

Environment:

Install the environment as the 'requirements.txt'.

pip install -i https://pypi.tuna.tsinghua.edu.cn/simple -r requirements.txt 

Besides, you should install pytorch according your CUDA version https://pytorch.org/get-started/locally/.

---

Command:

Check the run.sh script, and change hyperparameters accordingly. (when the dataset choise occurs, just choose the number of you dataset )

• 1 $ bash run.sh setup to run filternet.py 9_selective_search
• 2 $ bash run.sh patch to run patch_filter.py patchnet_bn_rerun.py
• 3 $ bash run.sh object to run saliencynet.py CAM.py objectnet_bn_rerun.py
• 4 $ bash run.sh part to run filterpart.py align_part.py partnet_bn_rerun.py
• 5 $ bash run.sh fusion to run fusion_predict.py

or just run $ bash run.sh all, the final result is in /log/fusion_predict.log