HanzhouLiu / DeblurDiNAT

Official implementation of the paper "DeblurDiNAT: A Generalizable Transformer for Perceptual Image Deblurring".
https://arxiv.org/abs/2403.13163
Other
29 stars 1 forks source link

How to test model on our testing images #1

Closed ghost closed 3 months ago

ghost commented 6 months ago

I am using the pretrained weights DeblurDiNATL.pth to test my own testing images. I have placed the pretrained model in the folder i.e. /results/DeblurDiNATL/models/DeblurDiNATL.pth. I have kept my testing images in the folder i.e. /DeblurDiNAT/input_images. I am ruuning script through: python predict_GoPro_test_results.py --job_name DeblurDiNATL --weight_name DeblurDiNATL.pth --blur_path /DeblurDiNAT/input_images. But I am encountering an error. The error is: Traceback (most recent call last): File "predict_GoPro_test_results.py", line 49, in for img_name in os.listdir(blur_path + '/' + file): NotADirectoryError: [Errno 20] Not a directory: '/home/hypersoft-gpu/Mahrukh_Ali_Khan/test/DeblurDiNAT/input_images/test6.jpg' Although the image having name test6.jpg is present in the respective folder

HanzhouLiu commented 6 months ago

Thanks for your attention to our work. I believe that error arises because '/home/hypersoft-gpu/Mahrukh_Ali_Khan/test/DeblurDiNAT/input_images/test6.jpg' is a file name instead of a directory. Our GoPro test structure looks like: Folder/SubFolder/Img.png. You can just change the for loop starting from line 49. Or you can move your test images into a sub directory so that the GoPro test codes work.

We will updated the other test codes, training codes, test results in future. Please stay tuned.