search

alterzero / DBPN-Pytorch

The project is an official implement of our CVPR2018 paper "Deep Back-Projection Networks for Super-Resolution" (Winner of NTIRE2018 and PIRM2018)
https://alterzero.github.io/projects/DBPN.html
MIT License
565 stars 164 forks source link

why i can't reproduce the effect with provided trained model #19

Closed liuiyilang closed 5 years ago

liuiyilang commented 6 years ago

I try to test the model on Set 5 X4 with provided trained models.When i calculate the psnr in RGB channel i got psnr =30.xxx which seems not right.When i calculate the psnr in Y channel,i got psnr = 31.xxx. Both of them seems lower than psnr presented in the paper(32.47)。Is there anyone got the results presented in the paper?

qingchuanhuajuan commented 5 years ago

Run python3 eval.py, no SR picture. Only the following sentences, can you achieve it? What do I need to change?

Namespace(chop_forward=True, gpu_mode=True, gpus=1, input_dir='Input', model='models/DBPN_x4.pth', model_type='DBPNL', output='Results/', seed=123, self_ensemble=True, testBatchSize=1, test_dataset='Set5_LR_x4', threads=1, upscale_factor=4) ===> Loading datasets ===> Building model Pre-trained SR model is loaded

qingchuanhuajuan commented 5 years ago

There are several models in the paper. Perhaps the X4 model that the author gives is not the best one. just guess.

iRmantou commented 5 years ago

@liuiyilang when i just use trained model weights "DBPN_x2.pth", I get wrong result, the result picture is green, did u get normall result?

liuiyilang commented 5 years ago

@liuiyilang when i just use trained model weights "DBPN_x2.pth", I get wrong result, the result picture is green, did u get normall result?

I got the normal results ,but seems the model doesn't performs that good with psnr.

liuiyilang commented 5 years ago

Run python3 eval.py, no SR picture. Only the following sentences, can you achieve it? What do I need to change?

Namespace(chop_forward=True, gpu_mode=True, gpus=1, input_dir='Input', model='models/DBPN_x4.pth', model_type='DBPNL', output='Results/', seed=123, self_ensemble=True, testBatchSize=1, test_dataset='Set5_LR_x4', threads=1, upscale_factor=4) ===> Loading datasets ===> Building model Pre-trained SR model is loaded

Sorry i can't help you,because i use my own dataloader and test code,while i just use dbpn.py and base_networks.py provided.

iRmantou commented 5 years ago

@liuiyilang thx, I have fixed it

yangyingni commented 5 years ago

I try to test the model on Set 5 X4 with provided trained models.When i calculate the psnr in RGB channel i got psnr =30.xxx which seems not right.When i calculate the psnr in Y channel,i got psnr = 31.xxx. Both of them seems lower than psnr presented in the paper(32.47)。Is there anyone got the results presented in the paper?

Have you dealt with it?And what's the PSNR of your result?Thank you.