Convert WAI logo to base64 for embedding in HTML. Returns: Base64 encoded logo string or None

Get model, loading if necessary.

Perform reconstruction using the already-created target_dir/images. Args: target_dir: Directory containing images

Create a new 'target_dir' + 'images' subfolder, and place user-uploaded images or extracted frames from video into it. Args:

Convert batched vectors (xyz) to (xyz0).

Initialize the DepthAnything3 model. Args: device: Device to load the model on

Compute the least-squares intersection of rays. Uses the math from here: https://math.stackexchange.com/a/1762491/286022

Compute least squares scale factor s such that a ≈ s * b. Args: a: First tensor b: Second tensor eps: Small epsilon

Loads config file. Args: path (str): path to config file. default_path (str, optional): whether to use default path. Default

Load DTU depth validity masks. Args: mask_files: List of paths to mask images Returns: Boolean arra

Load model

Main entry point for gallery server.

Main function to run the application.

Command-line interface for metrics printing.

Merge all PLY files in a directory into one file (without loading into memory) Args: - input_dir: Input directory containing multiple '{

Normalize the point vectors

Run the Levenberg Marquardt algorithm

Convert pixel space points to camera space points. Args: pixel_space_points (torch.Tensor): Pixel space points with shape (h, w, 2)

Populate the depth and measure tabs with processed data. Args: processed_data: Processed data dictionary Return

Convert a pose encoding back to camera extrinsics and intrinsics.

Args: chunk1_depth: (n1, h, w) chunk1_conf: (n1, h, w) chunk2_depth: (n2, h, w) chunk2_conf: (n2, h, w) m

Print comparison table for multiple evaluation runs. Args: metrics_list: List of metrics dictionaries labels

Process loop_list and return chunk indices and frame ranges for each (idx1, idx2) pair. chunk_index: List of (begin_idx, end_idx) tuples.

Convert pypose Sim3 to numpy s,R,t

Convert quaternion (WXYZ order) to rotation matrix. Args: qvec: Quaternion as [w, x, y, z] Returns: 3x3 rotation matrix

Batch version of RANSAC weighted Homography estimation (supports batch). Input: src_pts: (B, N, 2) dst_pts: (B, N, 2)

Read DTU camera file containing extrinsics and intrinsics. Args: filename: Path to camera text file Returns:

Read DTU camera file containing extrinsics and intrinsics. Args: filename: Path to camera text file Returns:

see: src/colmap/scene/reconstruction.cc void Reconstruction::WriteCamerasBinary(const std::string& path) void Reconstruction::Rea

see: src/colmap/scene/reconstruction.cc void Reconstruction::WriteCamerasText(const std::string& path) void Reconstruction::ReadC

see: src/colmap/scene/reconstruction.cc void Reconstruction::ReadImagesBinary(const std::string& path) void Reconstruction::Write

see: src/colmap/scene/reconstruction.cc void Reconstruction::ReadImagesText(const std::string& path) void Reconstruction::WriteIm

see: src/colmap/scene/reconstruction.cc void Reconstruction::ReadPoints3DBinary(const std::string& path) void Reconstruction::Wri

see: src/colmap/scene/reconstruction.cc void Reconstruction::ReadPoints3DText(const std::string& path) void Reconstruction::Write

data_list: [(67, (3386, 3406), 48, (2435, 2455)), ...]

Rendering stabilized camera path. poses: [batch, 4, 4] or [batch, 3, 4], return: smooth path: [batch 4 4]

cam2world: [batch, 4, 4], intr_normed: [batch, 3, 3]

Restore original view order after processing. Args: x: Reordered tensor of shape (B, S, ...) b_idx: Original reference v

src: (Nx3) tgt: (Nx3) init_weights: (N,)

Calculate rotation angle error between ground truth and predicted rotations. Args: rot_gt: Ground truth rotation matrices ro

Run complete loop detection pipeline

Run backend inference

Run DepthAnything3 model inference on images. Args: target_dir: Directory containing images filter_black_bg:

Run local inference

Get normalized (range 0 to 1) coordinates and integer indices for an image.

Sample tensor elements for quantile computation to reduce memory usage. Args: tensor: Input tensor to sample max_samples: Ma

Save camera poses from all chunks to txt and ply files - txt file: Each line contains a 4x4 C2W matrix flattened into 16 numbers

Save loop detection results to file

Save the current reconstruction results to the gallery directory. Args: target_dir: Source directory containing reconstruction resul

Compute rotation and translation errors between predicted and ground truth poses. Args: pred_se3: Predicted SE(3) transformations

Select a reference view from multiple views using the specified strategy. Args: x: Input tensor of shape (B, S, N, C) where

Solve linear system A * delta = b, supports submatrix solving

Convert a unit quaternion to a standard form: one in which the real part is non negative. Args: quaternions: Quaternions with re

Convert a unit quaternion to a standard form: one in which the real part is non negative. Args: quaternions: Quaternions with re

Transform points from 3D camera coordinates to 3D world coordinates.

Apply a rigid-body transformation to points or vectors.

frames: (T,H,W,D) or list of (H,W,D) returns: (T,H,W,3)

Calculate translation angle error between ground truth and predicted translations. Args: tvec_gt: Ground truth translation vectors

Unproject 2D camera coordinates with the given Z values.