[CPR] An error occurred while running code

[ZiMo-Chen]

When I execute the code [cmd 0], I find a problem, as shown below. I found that the implementation of evaluation seems to be missing in the code. Could you provide it? I would appreciate it if you could.

python=3.7 cuda=11.6 memory size=40G A100

[yinglang]

Hi, thanks for this report. this is due to some files(TOV_mmdetection/mmdet/models/detectors/base.py and TOV_mmdetection/mmdet/models/detectors/single_stage.py) are not updated.
These problem should been solved by replacing the files with the latest code.

[ZiMo-Chen]

Thank you for your reply

[ZiMo-Chen]

It's still the same problem

[yinglang]

Thanks for feedback. The problem can be solved by add **kwargs in the functions‘ defination just like this

[ZiMo-Chen]

Excuse me again, the problem is still there. And I can't find the function "simple_test_bboxes" where is defined.

[yinglang]

Sorry for that, I forget there is another function defination should be change, you can check there

[ZiMo-Chen]

Excuse me again. There seems to be some code missing in this file eval_hooks.py

[yinglang]

Thanks again. This problem have been solved by replace the TOV_mmdetection/mmdet/apis/train.py as the commit told.

[ZiMo-Chen]

https://github.com/ucas-vg/P2BNet/blob/main/TOV_mmdetection/mmdet/core/evaluation/eval_hooks.py
Can I modify eval_hook here according to eval_hooks in P2B

[yinglang]

I have updatd the eval_hooks.py. And eval_hooks.py in P2B should work too.

[ZiMo-Chen]

Excuse me again,I don't know why this error occurs

[yinglang]

Does the problem still exists?

[ZiMo-Chen]

The problem remains

[yinglang]

Excuse me again,I don't know why this error occurs

Does this problem still exists?

[yinglang]

The problem remains

The problem have been solved by replacing TOV_mmdetection/mmdet/datasets/coco.py.

you can re-evaluate CPR by resuming from trained parameter with 'resume-from' args and then use result2ann.py to transform result to annotation.

# [cmd 0]
GPU=2 && CUDA_VISIBLE_DEVICES=0,1 PORT=10001 tools/dist_train.sh \
  configs2/DOTA//coarsepointv2/coarse_point_refine_r50_fpns4_1x_DOTA_1024.py 2 \
  --work-dir ../TOV_mmdetection_cache/work_dir/DOTA//coarsepointv2/noise_rg-0-0-0.25-0.25_1/coarse_point_refine_r50_fpns4_1x_DOTA_1024/loss0gt_r7_7_lr0.0001_1x_1b4g_s8/ \
  --cfg-options optimizer.lr=0.0001 data.samples_per_gpu=1 \
    model.bbox_head.strides=[8] model.neck.start_level=1 \
    model.bbox_head.refine_pts_extractor.pos_generator.radius=7 \
    model.bbox_head.refine_pts_extractor.neg_generator.radius=7  \
    model.bbox_head.train_pts_extractor.pos_generator.radius=7 \
    model.bbox_head.train_pts_extractor.neg_generator.radius=7 \
    evaluation.save_result_file=../TOV_mmdetection_cache/work_dir/DOTA//coarsepointv2/noise_rg-0-0-0.25-0.25_1/coarse_point_refine_r50_fpns4_1x_DOTA_1024/loss0gt_r7_7_lr0.0001_1x_1b4g_s8//latest_result_refine2_2_r7_7.json
--resume-from  ../TOV_mmdetection_cache/work_dir/DOTA//coarsepointv2/noise_rg-0-0-0.25-0.25_1/coarse_point_refine_r50_fpns4_1x_DOTA_1024/loss0gt_r7_7_lr0.0001_1x_1b4g_s8/latest.pth

# [cmd 1]
python exp/tools/result2ann.py --ori_ann data/dota/DOTA-split/trainsplit/noise_rg-0-0-0.25-0.25_1/pseuw16h16/DOTA_train_1024_coarse.json \
  --det_file ../TOV_mmdetection_cache/work_dir/DOTA//coarsepointv2/noise_rg-0-0-0.25-0.25_1/coarse_point_refine_r50_fpns4_1x_DOTA_1024/loss0gt_r7_7_lr0.0001_1x_1b4g_s8//latest_result_refine2_2_r7_7.json \
  --save_ann ../TOV_mmdetection_cache/work_dir/DOTA//coarsepointv2/noise_rg-0-0-0.25-0.25_1/coarse_point_refine_r50_fpns4_1x_DOTA_1024/loss0gt_r7_7_lr0.0001_1x_1b4g_s8//DOTA_train_1024_refine2_2_r7_7.json

[ZiMo-Chen]

Thanks, the problem is mostly solved