Update app.py

app.py

@@ -31,9 +31,9 @@ model = torch.quantization.quantize_dynamic(
 
 processor = DPTImageProcessor.from_pretrained("Intel/dpt-swinv2-tiny-256")
 
-color_map = torch.from_numpy(cv2.applyColorMap(np.arange(256, dtype=np.uint8), cv2.COLORMAP_INFERNO)).to(device)
+color_map = cv2.applyColorMap(np.arange(256, dtype=np.uint8), cv2.COLORMAP_INFERNO)
 
-input_tensor = torch.zeros((1, 3, 72, 128), dtype=torch.float32, device=device)
+input_tensor = torch.zeros((1, 3, 128, 128), dtype=torch.float32, device=device)
 
 def preprocess_image(image):
     return cv2.resize(image, (128, 72), interpolation=cv2.INTER_AREA).transpose(2, 0, 1).astype(np.float32) / 255.0
@@ -46,16 +46,14 @@ def process_frame(image):
     input_tensor = torch.from_numpy(preprocessed).unsqueeze(0).to(device)
     
     predicted_depth = model(input_tensor).predicted_depth
-    depth_map = predicted_depth.squeeze()
-    
-    # Normalize depth map to [0, 255] range
+    depth_map = predicted_depth.squeeze().cpu().numpy()
+    num_bins = 1000
+    depth_map = np.digitize(depth_map, bins=np.linspace(depth_map.min(), depth_map.max(), num_bins)) - 1
     depth_map = (depth_map - depth_map.min()) / (depth_map.max() - depth_map.min())
-    depth_map = (depth_map * 255).byte()
-    
-    # Apply color map
-    depth_map_colored = color_map[depth_map.long()]
+    depth_map = (depth_map * 255).astype(np.uint8)
+    depth_map_colored = cv2.applyColorMap(depth_map, cv2.COLORMAP_INFERNO)
     
-    return cv2.cvtColor(depth_map_colored.cpu().numpy(), cv2.COLOR_BGR2RGB)
+    return cv2.cvtColor(depth_map_colored, cv2.COLOR_BGR2RGB)
 
 interface = gr.Interface(
     fn=process_frame,