search

DavisMeee / LighTDiff

LighTDiff: Surgical Endoscopic Image Low-Light Enhancement with T-Diffusion
MIT License
17 stars 0 forks source link

请教问题 #3

Closed linyteating closed 2 months ago

linyteating commented 2 months ago

你好，不知道是不是因为图片输入路径的原因，导致我们生成的图片非常奇怪。看你们的文档并没有涉及到这个问题，想请教一下应该怎么解决。 dd41287b9c8b4fa3b5673078eea1b13 4438dc295a34d3c277f9fdc86163af4

linyteating commented 2 months ago

发邮件请教和申请加微信，都没有回复

DavisMeee commented 2 months ago

Sorry for the late respond, but I don't know what happened to this with a limited screenshot. I would like to know whether you have changed any codes from the original.

linyteating commented 2 months ago

你好 我没改过代码 请问上面那几个路径应该放什么内容？

DavisMeee commented 2 months ago

Let me use LOL dataset as an example: input_root : ./our485/low gt_root : ./our485/high

linyteating commented 2 months ago

![Uploading 2ffcf98970bd7c3945b1b1e522fd322.png…]() 你好 我发现C:\Users\Administrator\Desktop\LighTDiff-main\BasicSR-light\docs 这个路径下有详细的配置要求，但是似乎并没有关于Lightdiff这个模型的配置要求，BasicSR-light\docs这个文件夹里面也只有关于超分要用什么数据集，并没有涉及到Lightdiff要用什么数据集，所以想向你请教一下怎么解决，应该放什么数据集比较合适。我放了一组高低光照匹配的数据集，但是生成的图片是这样的失真图片，感到非常奇怪。

linyteating commented 2 months ago

![Uploading 2ffcf98970bd7c3945b1b1e522fd322.png…]()

linyteating commented 2 months ago

你好，我发现LighTDiff-main\LighTDiff\BasicSR\options\test只有这个文件夹里面有超分的预训练模型，但是LighTDiff-main\LighTDiff\lightdiff这个文件夹下面没有关于diffusion的预训练模型，是没有吗？

DavisMeee commented 2 months ago

Sorry, there's no pretraining lightdiff model in our light-BasicSR. Basic-SR only has pre-training weights for classic models.

DavisMeee commented 2 months ago

BTW your upload images are broken and I am afraid that I cannot help with these broken images. I don't know what happens, but I am sure my code is reproducible.

linyteating commented 2 months ago

LighTDiff: Surgical Endoscopic Image Low-Light Enhancement with T-Diffusion 你好，上面这篇论文是你们所编撰的，那既然没有lightdiff的预训练模型，那 你们是怎么实现代码的？

linyteating commented 2 months ago

方便拉我进你们的交流群吗？QQ群满了，微信群加了负责人，到现在都还没有通过。

DavisMeee commented 2 months ago

We have utilized the library BasicSR to implement lightdiff framework. We are still preparing it and the pre-trained model will be released later. This model is fully retrainable and takes only 4 hours to train with batch_size = 8.

linyteating commented 2 months ago

你好 LighTDiff-main\experiments\training_archived_20240901_173614\models 我想问一下为什么训练完这里面的model是空的，是我们哪里出错了吗？还是你们代码没有输出出来?

DavisMeee commented 2 months ago

Right, first I don’t know what really happened with these single description. Then this code is reproducible and already verified by the others( like previous issues, they all succeed reproduced lightdiff). It must be something wrong with your env or code even database. Please check your code and debug it by yourself. Frankly speaking, this is not in my charge, thank you for understanding.

linyteating commented 2 months ago

好的 谢谢