Update app.py

app.py

@@ -5,7 +5,6 @@ import numpy as np
 from torchvision import transforms
 from PIL import Image
 from transformers import DPTForDepthEstimation, DPTFeatureExtractor
-import torchvision.transforms.functional as F
 
 # Load depth estimation model
 model_name = "Intel/dpt-large"
@@ -15,39 +14,53 @@ depth_model.eval()
 
 def estimate_depth(image):
     """Estimate depth map from image."""
-    image = image.convert("RGB").resize((384, 384))  # Resize for model input
+    image = image.convert("RGB")
+    image = image.resize((384, 384))  # Resize for model input
     inputs = feature_extractor(images=image, return_tensors="pt")
     with torch.no_grad():
         outputs = depth_model(**inputs)
         depth = outputs.predicted_depth.squeeze().cpu().numpy()
-    depth = cv2.resize(depth, (image.width, image.height))  # Resize back
+    depth = cv2.resize(depth, (image.width, image.height))  # Resize back to original
     depth = (depth - depth.min()) / (depth.max() - depth.min()) * 255
     return depth.astype(np.uint8)
 
-def warp_design(cloth_img, design_img):
-    """Warp the design onto the clothing while preserving folds."""
+def blend_design(cloth_img, design_img):
+    """Blend design onto clothing naturally."""
     cloth_img = cloth_img.convert("RGB")
     design_img = design_img.convert("RGB")
     cloth_np = np.array(cloth_img)
     design_np = np.array(design_img)
+    
+    # Resize design to fit within clothing
     h, w, _ = cloth_np.shape
+    dh, dw, _ = design_np.shape
+    scale_factor = min(w / dw, h / dh) * 0.6  # Scale to 60% of clothing area
+    new_w, new_h = int(dw * scale_factor), int(dh * scale_factor)
+    design_np = cv2.resize(design_np, (new_w, new_h))
+    
+    # Convert design to grayscale and darken
+    design_gray = cv2.cvtColor(design_np, cv2.COLOR_RGB2GRAY)
+    design_np = cv2.cvtColor(design_gray, cv2.COLOR_GRAY2RGB)
+    design_np = cv2.convertScaleAbs(design_np, alpha=1.2, beta=-30)  # Increase contrast
     
-    # Estimate depth map
+    # Create a blank canvas and paste the resized design at the center
+    design_canvas = np.zeros_like(cloth_np)
+    x_offset = (w - new_w) // 2
+    y_offset = (h - new_h) // 2
+    design_canvas[y_offset:y_offset+new_h, x_offset:x_offset+new_w] = design_np
+    
+    # Estimate depth for fold detection
     depth_map = estimate_depth(cloth_img)
     depth_map = cv2.resize(depth_map, (w, h))
     
-    # Compute optical flow for warping
-    flow = cv2.calcOpticalFlowFarneback(depth_map, depth_map, None, 0.5, 3, 15, 3, 5, 1.2, 0)
-    flow_map = np.column_stack((flow[..., 0] + np.arange(w), flow[..., 1] + np.arange(h)[:, None]))
-    warped_design = cv2.remap(design_np, flow_map, None, cv2.INTER_LINEAR, borderMode=cv2.BORDER_REFLECT)
-    
-    # Blending
-    blended = cv2.addWeighted(cloth_np, 0.7, warped_design, 0.3, 0)
+    # Use Poisson blending for seamless integration
+    mask = (design_canvas > 0).astype(np.uint8) * 255
+    blended = cv2.seamlessClone(design_canvas, cloth_np, mask, (w//2, h//2), cv2.NORMAL_CLONE)
     
     return Image.fromarray(blended)
 
 def main(cloth, design):
-    return warp_design(cloth, design)
+    return blend_design(cloth, design)
 
 iface = gr.Interface(
     fn=main,