Training with U-Net

[clue3clue32]

Hi, I was trying to reproduce the training process of Unet with unet_arch.py under the basicsr/archs folder. I changed the parameter 'type' of 'network_g' to 'U_Net' and remove the 'img_size' in uformer_flare7kpp_baseline_option.yml. No errors were reported during the training process. The results of running test.py on checkpoints are as follows : net_g_5000: net_g_10000: net_g_25000: What led to the occurrence of this problem? Thank you!

[ykdai]

Have you ever tried to modify the config using U-Net based on 'uformer_flare7k_option.yml'? In my memory, using U-Net on flare7kpp's setting is a little bit difficult to converge.

[clue3clue32]

Thanks for your reply, the results based on uformer_flare7k_option.yml under the blend folder are like this:

[ykdai]

Have you ever checked the dataloader's output of your training data, or can you report the trainning log? The seems that the deflared image is just the black image.