Update app.py

app.py

@@ -576,20 +576,25 @@ def draw_mask(l, t, v, d, evt: gr.EventData):
 
 def findNormals():
     global depths
-    depth = cv2.imread(depths[frame_selected]).astype(np.float32)
-    normals = np.zeros((depth.shape[0], depth.shape[1], 3), dtype=np.float32)
-
-    for x in range(1, depth.shape[1] - 1):
-        for y in range(1, depth.shape[0] - 1):
-            t = np.array([x, y - 1, depth[y - 1, x]], dtype=np.float32)
-            l = np.array([x - 1, y, depth[y, x - 1]], dtype=np.float32)
-            c = np.array([x, y, depth[y, x]], dtype=np.float32)
-
-            d = np.cross(l - c, t - c)
-            n = d / np.linalg.norm(d)
-            normals[y, x] = n
-
-    return (normals * 255 + 127).astype(np.uint8)
+    d_im = cv2.imread(depths[frame_selected]).astype(np.uint8)
+    zy, zx = np.gradient(d_im)  
+    # You may also consider using Sobel to get a joint Gaussian smoothing and differentation
+    # to reduce noise
+    #zx = cv2.Sobel(d_im, cv2.CV_64F, 1, 0, ksize=5)     
+    #zy = cv2.Sobel(d_im, cv2.CV_64F, 0, 1, ksize=5)
+
+    normal = np.dstack((-zx, -zy, np.ones_like(d_im)))
+    n = np.linalg.norm(normal, axis=2)
+    normal[:, :, 0] /= n
+    normal[:, :, 1] /= n
+    normal[:, :, 2] /= n
+
+    # offset and rescale values to be in 0-255
+    normal += 1
+    normal /= 2
+    normal *= 255
+
+    return normal[:, :, ::-1]
 
 
 load_model="""