Closed oscarlorente closed 3 years ago
Basically, you need to project the head vertices to image space and obtain head bbox from the projected 2D head vertices.
You can check the code here, where pred_vertices_img
is already aligned to the image space. You can directly take the [x, y]
coordinates to obtain the projected 2D vertices. Since you already know how to select out head vertices, you just need to do that again to pred_vertices_img
to obtain the head vertices in image space and then obtain 2D head bbox.
That works like a charm, thank you very much!
@oscarlorente
@oscarlorente How did you get this picture and how to paste the RGB information into the generated mesh.
Hello, I'm using frankmocap to get the SMPL-X model of a man, but I also need to detect his head using the corresponding SMPL-X prediction. So far, I've been able to identify the vertices corresponding to the head (using the SMPL-X - FLAME correspondences from here) and plot it using your openGL renderer:
However, I'm interested in getting the head's bbox coordinates, that is, knowing the pixels on which the SMPL-X vertices are projected. I'm not sure how to get them from the renderer code, any ideas?