error:flag = is_cross_text([start_w, start_h], length, new_vertices[labels==1,:]) IndexError: too many indices for array

[pycoco]

[Gyann-z]

hi , Does your question solved? I had the same problem.

[SakuraRiven]

Sry for the late reply. More information would be helpful if you still need help...

[dagonglu]

Does your question solved? I had the same problem.

[SakuraRiven]

Maybe check cross_text is not convenient and simply ignoring the boundary text is also ok.

[dagonglu]

Yes, the problem lies in the data marked on the edge

[ZdsAlpha]

Finally solved this issue. The problem occurs, when there are images in your gt that contain zero bounding boxes. So the shape of vertices becomes one dimensional which was supposed to be (total_vertices, 8). Replace this line in dataset.py: new_vertices = np.zeros(vertices.shape) with new_vertices = np.zeros((vertices.shape[0], 8))