search

jbeomlee93 / AdvCAM

Anti-Adversarially Manipulated Attributions for Weakly and Semi-Supervised Semantic Segmentation (CVPR 2021)
MIT License
128 stars 16 forks source link

Performance Gap on Init Seed #16

Closed PlumedSerpent closed 3 years ago

PlumedSerpent commented 3 years ago

Hi, I ran your codes and apply all the hyper-parameters by default. But the init seed mIoU is only 48.75, which is reported as 55.6 in your paper. I wonder if i miss something? Here is my train command line: CUDA_VISIBLE_DEVICES=4,5,6,7 python run_sample.py --voc12_root Dataset/VOC2012_SEG_AUG/ --cam_out_dir result/cam_res50 --sem_seg_out_dir result/seg_res50 --ir_label_out_dir result/ir_label_res50 --irn_weights_name sess/res50_irn.pth --make_ins_seg_pass False --eval_ins_seg_pass False

And here is my init seed result:

image

jbeomlee93 commented 3 years ago

Hi,

To obtain the initial seed of AdvCAM, you should run "obtain_CAM_masking.py", not "run_sample.py".

Please refer to "get_seed_quality.sh"

https://github.com/jbeomlee93/AdvCAM/blob/fa08f0ad4c1f764f3ccaf36883c0ae43342d34c5/get_seed_quality.sh#L3-L4

Thanks.

PlumedSerpent commented 3 years ago

Hi,

To obtain the initial seed of AdvCAM, you should run "obtain_CAM_masking.py", not "run_sample.py".

Please refer to "get_seed_quality.sh"

https://github.com/jbeomlee93/AdvCAM/blob/fa08f0ad4c1f764f3ccaf36883c0ae43342d34c5/get_seed_quality.sh#L3-L4

Thanks.

Got it, thanks.