search

albrateanu / LYT-Net

LYT-Net: Lightweight YUV Transformer-based Network for Low-Light Image Enhancement
https://arxiv.org/abs/2401.15204
MIT License
70 stars 11 forks source link

train.py中的h5文件 #7

Closed xuehaohewangyu closed 7 months ago

xuehaohewangyu commented 7 months ago

Hi author, how do I get the .h5 file from my own trained network, it only generates experiments, and this file is just the h5 generated by different epochs, there is no overall parameter file, how can I solve it?

albrateanu commented 7 months ago

Hello, thanks for writing.

To train LYT-Net using your custom parameters, you will have to modify scripts/train.py, model/losses.py and model/arch.py as per your requirements.

We will provide a configuration where users can set their custom parameters with more ease, but until then, this is the only way to train on custom options.

xuehaohewangyu commented 7 months ago

Thank you very much for your answer, I see that you are providing the pre-training weights h5 file, so that means that if I train, the last h5 of the experiments is the best parameter for the whole training process?

albrateanu commented 7 months ago

The pretrained model weights are the best results we managed to obtain. And if you train the model, the .h5 will be saved at the best metrics at that current time.