Open AIBluefisher opened 1 year ago
Hi,
I read the code in projection.py. However, I wonder why we need to normalize the pixels after reprojecting 3D points to the image plane.
def normalize(self, pixel_locations, h, w): resize_factor = torch.tensor([w-1., h-1.]).to(pixel_locations.device)[None, None, :] normalized_pixel_locations = 2 * pixel_locations / resize_factor - 1. # [n_views, n_points, 2] return normalized_pixel_locations
i think it is used for querying feature.
feat_sampled = F.grid_sample(featmaps, normalized_pixel_locations, align_corners=True)
(ibrnet/projection.py:121)
Hi,
I read the code in projection.py. However, I wonder why we need to normalize the pixels after reprojecting 3D points to the image plane.