Unable to reproduce experimental results

krasserm / super-resolution

Tensorflow 2.x based implementation of EDSR, WDSR and SRGAN for single image super-resolution

Apache License 2.0

1.5k stars 352 forks source link

Unable to reproduce experimental results #12

Closed huziye closed 5 years ago

huziye commented 5 years ago

I can't reproduce the results of the experiment based on the instructions you provided, using your code and the same data set.

krasserm commented 5 years ago

Do you mind sharing the commands you ran and the results you got :wink: ?

huziye commented 5 years ago

python train.py --dataset ./DIV2K_BIN --outdir ./output --profile wdsr-a-16 --scale 2 python evaluate.py --dataset ./DIV2K_BIN -i ./output/dir/models -o eval.json Your result is 34.68,but I get Best PSNR = 34.6105 and Best PSNR = 34.5994 in my two repeated experiments.

krasserm commented 5 years ago

Thanks, try training with 400 or 500 epochs (default is 300). Maybe I forgot adding that to the documented command.

krasserm commented 5 years ago

Please re-open if you still have problems to reproduce