About LOLv2_real configuration

[Sttrr]

Hello, we use the config file in the folder 'options/train/LOLv2_real.yml' to retrain the model, but the results obtained are a little worse than the ones presented in the paper. I dont know what i did wrong, Can you give me some help? my dataset and config with the results are presented below:

[wangchx67]

Hello,

Apologies for the delayed response. I've cloned the repository and retrained the model locally, and it successfully achieves the similar results outlined in the paper. You may wish to experiment with setting the GT_size and LQ_size to 384. Additionally, the seed used for my reproduction is 7335, so you can try to set your manual_seed to 7335 as well.

Besides, you also can directly use our pre-trained model in ./pre-trained/lol2-real.pth.

Best regards,

[Sttrr]

Thans for your response,

I modified the GT_size and random seed as you suggested, and it did improve. However, it is still a little worse than the results outlined in the paper. My best results were around 20000 iterations and slightly lower thereafter. Is that normal? I want to use your network in my work so i hope to retrain it. Can you send me your training log? It will be useful for me. My Gmail address is [email protected]

[wangchx67]

please check your email

[Newbie-v]

Could you please send me a copy of the training logs? Thank you.My Gmail address is [email protected]

请检查您的电子邮件

[wangchx67]

check the email