search

chxy95 / HDRUNet

CVPR2021 Workshop - HDRUNet: Single Image HDR Reconstruction with Denoising and Dequantization.
MIT License
157 stars 28 forks source link

odd image dimensions cause error #6

Open coordxyz opened 3 years ago

coordxyz commented 3 years ago

FYI, the odd image dimensions cause error and it can be fixed by adding pre-padding or pre-truncating. Since the NTIRE2021 dataset is not available now, I test the code with SICE dataset but without having good output. Is there any parameter I can play with? Input: 205 Output: 205

chxy95 commented 3 years ago

@bhyzhao Hello, the dataset of NTIRE is still available. I have just checked it. For the above image, it looks like it hasn't been processed by tone-mapping properly. If there is a ground truth of the image, you can try to process the image by μ-law tone-mapping we provided for visualization.

YangGangZhiQi commented 1 year ago

@chxy95 Thanks for your great work. I am curious about if there is not a ground truth, Could the model be available for inference?

andre20000131 commented 1 year ago

@coordxyz hello,i want to know that your inference.your output picture of inference.i want to know that if you used the pre model or finetuned model?