Inference does not work correctly about dct-net.pytorch

[LeslieZhoa]

Hi @darkasevgen ! Thanks for your question. I fix the slice bug and upload a new model. You can try again!

[darkasevgen]

Can you tell me more about how gif is received ?
Did you apply resize to a certain resolution and run the model? Or can the inference be done at any resolution?

[LeslieZhoa]

Hi @darkasevgen , the gif is generated with the whole image size. just like

img = cv2.imread('')
img_h,img_w,_ = img.shape 
n_h,n_w = img_h // 8 * 8,img_w // 8 * 8
img = cv2.resize(img,(n_w,n_h))

the model can run at any resolution which can be divided with no remainder.
but the result will be slightly worse.

[darkasevgen]

Thank you for your reply.