Open an image or a depthmap with opencv-python.

Apply a geometric transformation to a list of 3-D points. H: 3x3 or 4x4 projection matrix (typically a Homography) p: numpy/torch/tuple of c

Invert a torch or numpy matrix

This function: - first downsizes the image with LANCZOS inteprolation, which is better than bilinear interpolation in

Args: - depthmap (HxW array): - camera_intrinsics: a 3x3 matrix - camera_pose: a 4x3 or 4x4 cam2world matrix Returns:

For the "metric-based" approach, we do either a global or per-trajectory scale, or a global Sim(3) alignment, then measure fraction of points

image has B x 3 x img_size x img_size do_mask: whether to perform masking or not return_all_blocks: if True, return the feat

Closed-form Sim(3) estimation via SVD, aligning A to B. Returns: s, R, t

Return a 3D point cloud corresponding to valid pixels of the depth map

Args: gt_poses: Tensor of shape (B, 4, 4) representing the ground truth poses pred_poses: Tensor of shape (B, 4, 4)

Reprojection method, for when the absolute depth is known: 1) estimate the camera focal using a robust estimator 2) reproject points

Convert flow into middlebury color code image :param flow: optical flow map :return: optical flow image in middlebury color

build a prediction head for the decoder

Generic method iterating through a base_path folder to find scenes.

Log current memory usage.

Args: quaternions (torch.Tensor): (*, 4) Returns: torch.Tensor: (*, 3, 3)

Predict in the correct aspect-ratio, then transpose the result in landscape and stack everything back together.

Args: yaw (torch.Tensor): (*) Returns: torch.Tensor: (*, 3, 3)

Init. Args: scale_factor (float): scaling mode (str): interpolation mode

return_all_blocks: if True, return the features at the end of every block instead of just the features from the l