Update app.py

app.py

@@ -139,7 +139,7 @@ def start_tryon(dict, garm_img, garment_des, is_checked, is_checked_crop, denois
     pipe.to(device)
     pipe.unet_encoder.to(device)
 
-    garm_img = garm_img.convert("RGB").resize((705,705))
+    garm_img = garm_img.convert("RGB").resize((768, 768))
     human_img_orig = dict["background"].convert("RGB")
 
     if is_checked_crop:
@@ -152,17 +152,17 @@ def start_tryon(dict, garm_img, garment_des, is_checked, is_checked_crop, denois
         bottom = (height + target_height) / 2
         cropped_img = human_img_orig.crop((left, top, right, bottom))
         crop_size = cropped_img.size
-        human_img = cropped_img.resize((705,705))
+        human_img = cropped_img.resize((768, 768))
     else:
-        human_img = human_img_orig.resize((705,705))
+        human_img = human_img_orig.resize((768, 768))
 
     if is_checked:
         keypoints = openpose_model(human_img.resize((384, 512)))
         model_parse, _ = parsing_model(human_img.resize((384, 512)))
         mask, mask_gray = get_mask_location('hd', categorie , model_parse, keypoints)
-        mask = mask.resize((705,705))
+        mask = mask.resize((768, 768))
     else:
-        mask = dict['layers'][0].convert("RGB").resize((705,705))#pil_to_binary_mask(dict['layers'][0].convert("RGB").resize((705,705)))
+        mask = dict['layers'][0].convert("RGB").resize((768, 768))#pil_to_binary_mask(dict['layers'][0].convert("RGB").resize((768, 768)))
     mask_gray = (1 - transforms.ToTensor()(mask)) * tensor_transfrom(human_img)
     mask_gray = to_pil_image((mask_gray + 1.0) / 2.0)
 
@@ -172,7 +172,7 @@ def start_tryon(dict, garm_img, garment_des, is_checked, is_checked_crop, denois
     args = apply_net.create_argument_parser().parse_args(('show', './configs/densepose_rcnn_R_50_FPN_s1x.yaml', './ckpt/densepose/model_final_162be9.pkl', 'dp_segm', '-v', '--opts', 'MODEL.DEVICE', 'cuda'))
     pose_img = args.func(args, human_img_arg)
     pose_img = pose_img[:, :, ::-1]
-    pose_img = Image.fromarray(pose_img).resize((705,705))
+    pose_img = Image.fromarray(pose_img).resize((768, 768))
 
     with torch.no_grad():
         with torch.cuda.amp.autocast():
@@ -226,10 +226,10 @@ def start_tryon(dict, garm_img, garment_des, is_checked, is_checked_crop, denois
                     cloth=garm_tensor.to(device, torch.float16),
                     mask_image=mask,
                     image=human_img,
-                    height=1024,
+                    height=768,
                     width=768,
-                    ip_adapter_image=garm_img.resize((705,705)),
-                    guidance_scale=2.0,
+                    ip_adapter_image=garm_img.resize((768, 768)),
+                    guidance_scale=6.0,
                 )[0]
 
     if is_checked_crop:
@@ -353,7 +353,7 @@ def generate_mask(human_img, categorie='upper_body'):
         keypoints = openpose_model(human_img_resized)
         model_parse, _ = parsing_model(human_img_resized)
         mask, mask_gray = get_mask_location('hd', categorie , model_parse, keypoints)
-        mask = mask.resize((705,705))
+        mask = mask.resize((768, 768))
 
         # Redimensionner le masque à la taille d'origine de l'image
         mask_resized = mask.resize(human_img.size)