Distributed evaluation for TapVid3D data. 1) Inference on each npz track 2) Evaluate using "metric-based" scaling: global & per-traj

Get the transformation matrix that maps the local rotation delta in 3D tangent space to the 4D space where the quaternion is embedded

Open and convert all images or videos in a list or folder to proper input format for DUSt3R.

Compute the loss (and optionally other results) for one batch with timing. Args: batch_data: Tuple of (view1_combined, view2_combined

Convert a 7D pose (translation + quaternion) to a 4x4 transformation matrix. Args: pose (torch.Tensor): Shape (..., 7), where th

Reads a trajectory or file list from a text file. File format: "stamp d1 d2 d3 ..." Returns: dict -- dictionary of (stamp, [dat

Args: x3d (Tensor): Shape (num_obj, num_pts, 3) x2d (Tensor): Shape (num_obj, num_pts, 2) w2d (Tensor): S

Transfer some variables to another device (i.e. GPU, CPU:torch, CPU:numpy). batch: list, tuple, dict of tensors or other things device: pyto

Visualize reconstruction results with both 2D frames and 3D pointclouds. Combines multiple frames in a single visualization. Args:

flow_utils._computeColor(flow): compute color codes for the flow field flow ---- PARAMETERS ---- flow: np.array of dimension (he

A helper function that computes the so3 exponential map and, apart from the rotation matrix, also returns intermediate variables that can

Convert a camera-to-world matrix to a tuple of translation and rotation input: c2w: 4x4 matrix output: tuple of translation and rota

Vectorized normal estimation using shared adjacency structure. Args: points (torch.Tensor): (B, N, 3) adjacency (torch.Tensor):

Compute 'overlapping' metrics based on a distance threshold between two point clouds.

Args: - depthmap (BxHxW array): - pseudo_focal: [B,H,W] ; [B,2,H,W] or [B,1,H,W] Returns: pointmap of absolute coordi

Optimized version of distance preservation loss using K-nearest neighbors in 3D space. Vectorized across time dimension to avoid for-loop.

Apply dynamic-static reweighting to confidence values based on trajectory movement. Args: conf1_ori: (B_subset, H, W) Original c

Estimate a global Sim(3) transformation that aligns A to B in a least-squares sense. That is, find s, R, t such that: B ~ s * R * A +

Args: x3d (torch.Tensor): Shape (*, n, 3) x2d (torch.Tensor): Shape (*, n, 2) w2d (torch.Tensor): Shape (*, n, 2)

embed_dim: output dimension for each position pos: a list of positions to be encoded: size (M,) out: (M, D)

grid_size: int of the grid height and width return: pos_embed: [grid_size*grid_size, embed_dim] or [n_cls_token+grid_size*grid_size, embe