Update app.py

app.py

@@ -24,7 +24,7 @@ os.makedirs(LR_path)
 os.makedirs(Ref_path)
 os.makedirs(Ref_path_T)
 os.makedirs('result')
-max_side = 512
+max_side = 1024
 def resize(img):
     #basewidth = max_side
     #wpercent = (basewidth/float(img.size[0]))
@@ -41,8 +41,8 @@ def resize(img):
     return img
   
 def inference(LR, Ref):
-    LR = resize(LR)
-    Ref = resize(Ref)
+    #LR = resize(LR)
+    #Ref = resize(Ref)
 
     LR.save(os.path.join(LR_path, '0000.png'))
     Ref.save(os.path.join(Ref_path, '0000.png'))
@@ -68,10 +68,10 @@ description="Demo application for Reference-based Video Super-Resolution (RefVSR
 
 article = "<p style='text-align: center'><b>To check the full capability of the module, we recommend to clone Github repository and run RefVSR models on videos using GPUs.</b></p><p style='text-align: center'>This demo runs on CPUs and only supports RefVSR for a single LR and Ref frame due to computational complexity. Hence, the model will not take advantage of temporal LR and Ref frames.</p><p style='text-align: center'>The model is the small-sized model trained with the proposed two-stage training strategy.</p><p style='text-align: center'>The frames will be resized so that the length of a longer side of the frames doesn't exceed 256 pixels.</p><p style='text-align: center'><a href='https://junyonglee.me/projects/RefVSR' target='_blank'>Project</a> | <a href='https://arxiv.org/abs/2203.14537' target='_blank'>arXiv</a> | <a href='https://github.com/codeslake/RefVSR' target='_blank'>Github</a></p>"
 
-LR = resize(Image.open('LR.png'))
-Ref = resize(Image.open('Ref.png'))
-LR.save('LR.png')
-Ref.save('Ref.png')
+#LR = resize(Image.open('LR.png'))
+#Ref = resize(Image.open('Ref.png'))
+#LR.save('LR.png')
+#Ref.save('Ref.png')
 
 examples=[[['LR.png'], ['Ref.png']]]