CLIFF
CLIFF copied to clipboard
about perspective camera translation
In losses.py
, it seems translation has been overlooked.
def perspective_projection(points,
rotation,
translation,
focal_length,
camera_center):
"""
This function computes the perspective projection of a set of points.
Input:
points (bs, N, 3): 3D points
rotation (bs, 3, 3): Camera rotation
translation (bs, 3): Camera translation
focal_length (bs,) or scalar: Focal length
camera_center (bs, 2): Camera center
"""
batch_size = points.shape[0]
K = torch.zeros([batch_size, 3, 3], device=points.device)
K[:, 0, 0] = focal_length
K[:, 1, 1] = focal_length
K[:, 2, 2] = 1.
K[:, :-1, -1] = camera_center
# Transform points
points = torch.einsum('bij,bkj->bki', rotation, points)
# Apply perspective distortion
projected_points = points / points[:, :, -1].unsqueeze(-1)
# Apply camera intrinsics
projected_points = torch.einsum('bij,bkj->bki', K, projected_points)
return projected_points[:, :, :-1]
translation
is not used. In perspective camera projection, points should add up translation