A question about the platform

[UtopiaHu]

Hi , Thanks for your work. The platform covers 13 popular deep learning-based LLIE methods, where the results of any inputs can be produced through a user-friendly web interface. So what training sets were these 13 methods trained with? Did you retrain them? Thank you for your reply.

[Li-Chongyi]

We directly use the pre-trained models to run the results for fair comparisons.

[UtopiaHu]

We directly use the pre-trained models to run the results for fair comparisons.

For fair comparisons, all the 13 methods should be trained using the same dataset. Isn't it?

[Li-Chongyi]

I mean we directly use the models released by the corresponding authors. We did not re-train these models as we are not sure whether or not the retrained model is correct due to the complex training strategies.

wideawaker @.***> 于2021年4月26日周一 下午4:01写道：

We directly use the pre-trained models to run the results for fair comparisons.

For fair comparisons, all the 13 methods should be trained using the same dataset. Isn't it?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/Li-Chongyi/Lighting-the-Darkness-in-the-Deep-Learning-Era-Open/issues/2#issuecomment-826604848, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIEVQEJWGBMCPAGEEFCVQU3TKUMWHANCNFSM43RCUDMA .

[UtopiaHu]

I mean we directly use the models released by the corresponding authors. We did not re-train these models as we are not sure whether or not the retrained model is correct due to the complex training strategies. wideawaker @.***> 于2021年4月26日周一 下午4:01写道： … We directly use the pre-trained models to run the results for fair comparisons. For fair comparisons, all the 13 methods should be trained using the same dataset. Isn't it? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIEVQEJWGBMCPAGEEFCVQU3TKUMWHANCNFSM43RCUDMA .

Thank you for your reply. So can the enhanced results of the platform be downloaded directly?

[Li-Chongyi]

We are afraid of the crush of GPU clusters. So, we may not provide the function of download for now. But, we are figuring out how to solve this issue. We hope this platform could provide the download function soon.

wideawaker @.***> 于2021年4月26日周一 下午5:45写道：

I mean we directly use the models released by the corresponding authors. We did not re-train these models as we are not sure whether or not the retrained model is correct due to the complex training strategies. wideawaker @.***> 于2021年4月26日周一 下午4:01写道： … <#m3081080571897691661> We directly use the pre-trained models to run the results for fair comparisons. For fair comparisons, all the 13 methods should be trained using the same dataset. Isn't it? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2 (comment) https://github.com/Li-Chongyi/Lighting-the-Darkness-in-the-Deep-Learning-Era-Open/issues/2#issuecomment-826604848>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIEVQEJWGBMCPAGEEFCVQU3TKUMWHANCNFSM43RCUDMA .

Thank you for your reply. So can the enhanced results of the platform be downloaded directly?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/Li-Chongyi/Lighting-the-Darkness-in-the-Deep-Learning-Era-Open/issues/2#issuecomment-826686209, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIEVQEK7SHQW6VKLJFY3BMLTKUY43ANCNFSM43RCUDMA .

[Li-Chongyi]

Our LoLi-Platform supports the function of download. Please right click and save the figure.

[DOU-AL]

I mean we directly use the models released by the corresponding authors. We did not re-train these models as we are not sure whether or not the retrained model is correct due to the complex training strategies. wideawaker @.***> 于2021年4月26日周一 下午4:01写道： … We directly use the pre-trained models to run the results for fair comparisons. For fair comparisons, all the 13 methods should be trained using the same dataset. Isn't it? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIEVQEJWGBMCPAGEEFCVQU3TKUMWHANCNFSM43RCUDMA .

Dear professor: Thanks for your work. The platform is great! But when I use one of the methods(Kind++) to run my image with its pre-trained models ,the result is not as well as the platform shows.So,can you share the Kind++'s pre-trained models the platform used?Thanks again and have a nice day!

[Li-Chongyi]

Due to the limited GPU size, we have to downsample the input if the platform comes to out of memory. So, the difference may be caused by resize operation. The results produced by the platform cannot be used for quantitative comparison. We just show the visual results on the platform.

DOU-AL @.***> 于2023年5月18日周四 14:20写道：

I mean we directly use the models released by the corresponding authors. We did not re-train these models as we are not sure whether or not the retrained model is correct due to the complex training strategies. wideawaker @.***> 于2021年4月26日周一 下午4:01写道： … <#m2308466213650788865> We directly use the pre-trained models to run the results for fair comparisons. For fair comparisons, all the 13 methods should be trained using the same dataset. Isn't it? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2 (comment) https://github.com/Li-Chongyi/Lighting-the-Darkness-in-the-Deep-Learning-Era-Open/issues/2#issuecomment-826604848>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIEVQEJWGBMCPAGEEFCVQU3TKUMWHANCNFSM43RCUDMA .

Dear professor: Thanks for your work. The platform is great! But when I use one of the methods(Kind++) to run my image with its pre-trained models ,the result is not as well as the platform shows.So,can you share the Kind++'s pre-trained models the platform used?Thanks again and have a nice day!

— Reply to this email directly, view it on GitHub https://github.com/Li-Chongyi/Lighting-the-Darkness-in-the-Deep-Learning-Era-Open/issues/2#issuecomment-1552503490, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIEVQEJSXJHBYZA75MCJY4LXGW5RBANCNFSM43RCUDMA . You are receiving this because you commented.Message ID: <Li-Chongyi/Lighting-the-Darkness-in-the-Deep-Learning-Era-Open/issues/2/1552503490 @github.com>