GuoLanqing comments

Results 12 comments of


                                            GuoLanqing

content discriminator training vs other parts training

Hi, have you solved this problem? I am also confused about this training setting.

Issue in torch 1.7.1

Hi, I meet the same question. The scale of your input image is [0,1]?

Dataset download link is broken

same problem. Thanks.

Some recent papers on shadow detection & dataset

Thanks for your suggestion. This paper has been included in `Shadow Generation` part.

what is ‘el‘ and ‘de‘ ，for example pixelwise_a3c_el and pixelwise_a3c_de

‘el‘ and ‘de‘ denote the enlightening and denoising. The 'de' version is designed for enhancing those low-light images with obvious noise corruptions, while 'el' is for enhancing those high-quality low-light...

Some files do not exist

This is the data path file. It contains the list of training or testing data paths. I put the example in ` ReLLIE/data/'.

got unpleasant result

Hi, same problem for me. Could you please provide the pretrained model? Many thanks!

ground truth image availability or workaround

Maybe you can try some unsupervised methods or use the adversarial loss to finetune the generator.

training time issue

Please refer to #25

Training iteration, epochs and batch_size for reproduce

Sorry for replying late. Please try `batchsize=16`, `num epochs=5000`, and `lr=3e-05`. The performance will be slightly improved training more epochs. If you want to train on your customised dataset, we...