wdzhao123 / APL

ECCV2022: United Defocus Blur Detection and Deblurring via Adversarial Promoting Learning
16 stars 3 forks source link

APL in PyTorch

Implementation of "United Defocus Blur Detection and Deblurring via Adversarial Promoting Learning" in PyTorch

Datasets

Download and unzip datasets from https://github.com/shangcai1/SG [1] to "./dataset". Add testset of DP datasets from https://ln2.sync.com/dl/c45358c50/r7kpybwk-xw8hhszh-qkj249ap-y8k2344d/view/default/10770664840008 [2] to "./dataset/test/DP".

Test

You can use the following command to test:

python test.py --image_path TEST_DATA_PATH --result_save_path RESULT_IMAGE_PATH

You can use the following model to output results directly.

Here are our parameters:

baidu link: https://pan.baidu.com/s/1sAbhPioPCLrsAid1W8UMAg?pwd=t8hq password: t8hq

google drive: https://drive.google.com/drive/folders/1lPcoIY-lqKAvsfvuYUgB9mmM6YKBy2h8?usp=sharing

Put "DBD.pth" and "deblur.pth" in "./saved_models".

Train

You can use the following command to train:

python train.py --data_root TRAIN_DATA_PATH

Eval

DBD

If you want to use Fmax and MAE to evaluate the results, you can run the following code in MATLAB. It shows the PR curve and F-measure curve at the same time.

./evaluate_dbd/evaluate.m

Deblurring

If you want to use PSNR, SSIM and MAE to evaluate the result, use the following code:

python evaluate.py --image_save_path RESULT_IMAGE_PATH --test_gt_path GT_PATH

References

[1] Zhao, W., Shang, C., Lu, H.: Self-generated defocus blur detection via dual adversarial discriminators. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[2] Abuolaim, A., Brown, M.S.: Defocus deblurring using dual-pixel data. In: European Conference on Computer Vision.