search

flyywh / Recurrent-Multi-Frame-Deraining

23 stars 5 forks source link

the performance of the released pertrained model #4

Open hcy96 opened 2 years ago

hcy96 commented 2 years ago

Hi, I test the released pre-trained model (w gan )on the RainSynAll100 testset.

The performance is lower (22.85, 0.85) than the result in the paper (25.14,0.91).

Could you please give me some suggestions to reproduce the result in the paper?

flyywh commented 2 years ago

We use wo_gan version for the metric evaluation and w_gan for the real-video testing. Thanks!

mapleTan1113 commented 2 years ago

老哥问我想问数据集的存放路径要怎么搞，代码里训练时用了两个数据集是怎么回事 Recurrent_multi_frame_deraining_wo_gan/train_video_rain.py，line228左右的 data_loader = utils.create_data_loader(train_dataset, opts, "train") data_haze_loader = utils.create_data_loader(train_haze_dataset, opts, "train")