Update app.py

app.py

@@ -42,38 +42,29 @@ def get_points_on_a_grid(
     device: Optional[torch.device] = torch.device("cpu"),
 ):
     r"""Get a grid of points covering a rectangular region
-
     `get_points_on_a_grid(size, extent)` generates a :attr:`size` by
     :attr:`size` grid fo points distributed to cover a rectangular area
     specified by `extent`.
-
     The `extent` is a pair of integer :math:`(H,W)` specifying the height
     and width of the rectangle.
-
     Optionally, the :attr:`center` can be specified as a pair :math:`(c_y,c_x)`
     specifying the vertical and horizontal center coordinates. The center
     defaults to the middle of the extent.
-
     Points are distributed uniformly within the rectangle leaving a margin
     :math:`m=W/64` from the border.
-
     It returns a :math:`(1, \text{size} \times \text{size}, 2)` tensor of
     points :math:`P_{ij}=(x_i, y_i)` where
-
     .. math::
         P_{ij} = \left(
              c_x + m -\frac{W}{2} + \frac{W - 2m}{\text{size} - 1}\, j,~
              c_y + m -\frac{H}{2} + \frac{H - 2m}{\text{size} - 1}\, i
         \right)
-
     Points are returned in row-major order.
-
     Args:
         size (int): grid size.
         extent (tuple): height and with of the grid extent.
         center (tuple, optional): grid center.
         device (str, optional): Defaults to `"cpu"`.
-
     Returns:
         Tensor: grid.
     """
@@ -100,13 +91,11 @@ def paint_point_track(
     colormap: Optional[List[Tuple[int, int, int]]] = None,
 ) -> np.ndarray:
   """Converts a sequence of points to color code video.
-
   Args:
     frames: [num_frames, height, width, 3], np.uint8, [0, 255]
     point_tracks: [num_points, num_frames, 2], np.float32, [0, width / height]
     visibles: [num_points, num_frames], bool
     colormap: colormap for points, each point has a different RGB color.
-
   Returns:
     video: [num_frames, height, width, 3], np.uint8, [0, 255]
   """
@@ -399,7 +388,7 @@ def track(
     # torch.cat([pred_tracks[:,:1],pred_tracks[:,1:]],dim=1)
     pred_tracks = (pred_tracks * torch.tensor([video_preview.shape[2], video_preview.shape[1]]).to(device) / torch.tensor([VIDEO_INPUT_RESO[1], VIDEO_INPUT_RESO[0]]).to(device))
     
-    vis = Visualizer(save_dir="./saved_videos", pad_value=0, linewidth=3, tracks_leave_trace=0)
+    vis = Visualizer(save_dir="./saved_videos", pad_value=0, linewidth=2, tracks_leave_trace=0)
     # segm_mask = torch.zeros(queries.shape[1])
     # segm_mask[:num_tracks] = 1
     # print('segm_mask',segm_mask.shape, segm_mask)