tqdm but with parallel execution. Will essentially return res = [ function(arg) # default function(*arg) # if star_args is

imgs: (B, 3, H, W) x: (B, L, patch_size**2 *3)

Visualizes 3D point trajectories for both GT and Pred tracks. Args: - gt_points (T, N, 3 np.array): Ground truth 3D points (world coordi

extract 3D points/confidence from prediction head output

Args: x3d (Tensor): Shape (*, n, 3) pose (Tensor): Shape (*, 4 or 7) out_jac (bool | Tensor): Shape (*, n

Project 3d points to 2d image plane.

extract confidence from prediction head output

extract 3D points from prediction head output

Convert the depth map's value to another camera pose. input: src_R: rotation matrix of source camera src_t: translation vecto

Sample a random viewpoint close to an existing one, using the navmesh and a reference observed point.

Saves the depth map for each camera view. The depth is calculated by transforming self.pts3d_j[i] (points in a model space) i

Save prediction results as NPY files for later re-evaluation. Args: pred_tracks: Tensor or numpy array of predicted 3D tracks

Scale predicted trajectories to align with ground truth trajectories by computing distances from each point to image center and optimizing a

Args: - screenspace ptc (Nx3 array): - camera_intrinsics: a 3x3 matrix Returns: pointmap of absolute coordinates (HxW

Read camera data, return (M,N) tuple. M is the intrinsic matrix, N is the extrinsic matrix, so that x = M*N*X, with x being a point in h

Convert extrinsic matrix E_old from old camera coordinate system (X left, Y up, Z inward) to extrinsic matrix E_new in OpenCV coordinate syst

Perform trajectory evaluation in different modes. Args: args: Command line arguments model: The model to evaluate

Perform Test-Time Adaptation (TTA) evaluation on test sequences. Args: args: Command line arguments model_without_ddp: T

Convert translation and quaternion to 4x4 extrinsic matrix.

Visualizes and saves the 2D track visualization on the video frames with alpha blending for track traces. Args: - imgs (list of T PIL.Im

Visualizes the results and saves two videos: 1. 2D track visualization (on video frames) 2. 3D track visualization (in world coordinates)

writeFlowFile(flow,<FILENAME>) write flow to the file <FILENAME>. if <FILENAME> does not exist, an IOError is raised. if <FILENAME> does