search

tengyu1998 / SCI

[CVPR 2022 Oral] This is the official code for the paper "Toward Fast, Flexible, and Robust Low-Light Image Enhancement".
MIT License
126 stars 69 forks source link

在其他光照条件图像上的结果 #51

Open Powerfulidot opened 2 days ago

Powerfulidot commented 2 days ago

直接用官方训练好的权重对过曝图像进行预测的话，效果不好，看起来更过曝了。请问这个模型除了低光照外，能用于过曝图像吗？如果可以的话，网络的损失或者其他部分是否需要修改什么呢？谢谢

the result is bad when i straightly use official pretrained weights to predict over-exposed pics and they looks more exposed. so i wonder can this model be used on over-exposed pics? if i does, do i need to modify the loss or other part of the network? thanks!

Powerfulidot commented 1 day ago

简单来说，我能通过对模型做某些修改来使模型可以进行过曝图像增强吗？ @tengyu1998

briefly saying, can i apply some modification to the model so that it can enhance over-exposed pics?