search

Tencent / Real-SR

Real-World Super-Resolution via Kernel Estimation and Noise Injection
Apache License 2.0
775 stars 118 forks source link

How to train x2 model? #15

Open yangshunDragon opened 3 years ago

yangshunDragon commented 3 years ago

Hello, I use kernelGAN to generate x4 kernels. If I want to train x2 model, should I re-generate x2 kernels with kernelGAN with --X2 parameters? or could I prepare the datapair by x4 kernels? What's more, could I finetune x2 model from pretrained RRDB_PSNR_x4.pth model ?

fourPieces0927 commented 3 years ago

Hello, I use kernelGAN to generate x4 kernels. If I want to train x2 model, should I re-generate x2 kernels with kernelGAN with --X2 parameters? or could I prepare the datapair by x4 kernels? What's more, could I finetune x2 model from pretrained RRDB_PSNR_x4.pth model ?

hello, do you know how to train x2 model now ??

syfbme commented 3 years ago

According to the source code, remove "--X4" in "CUDA_VISIBLE_DEVICES=4,5,6,7 python3 train.py --X4 --input-dir SOURCE_PATH" will meet your expectation