search

idealo / image-super-resolution

🔎 Super-scale your images and run experiments with Residual Dense and Adversarial Networks.
https://idealo.github.io/image-super-resolution/
Apache License 2.0
4.56k stars 755 forks source link

unreproducible results as introduction #203

Open SE2AI opened 3 years ago

SE2AI commented 3 years ago

dear authors,

I try to use the demo code and the baboon image in your introduction as below:

import numpy as np
from PIL import Image
from ISR.models import RDN
import matplotlib.pyplot as plt

img = Image.open('./input/sx.jpg')
lr_img = np.array(img)

rdn = RDN(weights='psnr-large')
sr_img = rdn.predict(lr_img, by_patch_of_size=50)
res = Image.fromarray(sr_img)
plt.imshow(res)
plt.show()

I test different modes of weight with psnr-large/psnr-small/noise-cancel. However, none of these could generate your result. Same problem also mentioned in #186

I think I must miss something, is there any suggestion?

regards

cfrancesco commented 3 years ago

Hi, are you able to reproduce the results on colab?

SE2AI commented 3 years ago

@cfrancesco no, it didn't work, the result seems not change.

cfrancesco commented 3 years ago

Could you post a screenshot of the results?

SE2AI commented 3 years ago

@cfrancesco sure, both the input and output like the image here https://github.com/idealo/image-super-resolution/blob/master/figures/baboon-compare.png, the output may be slightly different from the input, but one can not figure out with eyes' observation.