Zhangkai NI
Zhangkai NI
Please check if the size of your test image is a multiple of 2^4 (16).
You should crop all test images to a multiple of 16. Besides, if all test images have a uniform size, the ```val_batch_size``` can be set to be greater than 1,...
For the MIT-Adobe FiveK dataset, I directly used Lightroom to decode the images to png format and resized the short side of the images to 512 resolution. The 'low-quality' images...
@ComingPeopleHW Both the MIT-Adobe FiveK dataset and Flickr dataset used the 'identity loss'. If you have any further questions, feel free to contact me via email ([email protected]).
Yes, expertC stands for label, and original stands for raw. In order to generate unpaired training data, the subset is randomly divided into three partitions: 1) the first partition has...
yes, you can test arbitrary size if you comment line 97 in data_loader.py, but the size of your test image should be a multiple of 2^4 (16).
以一定的概率将之前增强的照片输入到辨别器，具体可以参考文章“Learning from Simulated and Unsupervised Images through Adversarial Training”的Section 2.3
我没有测试该模型在LOL数据集上的性能，如果你感兴趣可以对loss以及模型结构进行相应的修改然后测试一下在LOL数据集上的性能