Registry that ensures all datasets are auto-discovered and imported on first use.

nn.Module wrapper around Tensor.permute for cleaner nn.Sequential usage.

Main application class for Depth Anything 3 Gradio app.

Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).

Prepares a batch of images for model inference. This processor converts a list of image file paths into a single, model-ready tensor. The proc

Loop detector class for detecting loop closures in image sequences

Beautiful tabular metrics printer with color support. Features: - Colorized best/worst values - Grouped by dataset and evaluation mo

Lightweight residual convolution block for fusion

Loop closure optimizer for sequences of Sim3 transformations Input: - sequential_transforms: List[Tuple[float, np.ndarray, np.ndarray]]

Depth Anything 3 main API class. This class provides a high-level interface for depth estimation using Depth Anything 3. It supports bot

Main evaluation orchestrator for DepthAnything3 benchmarks. Usage: evaluator = Evaluator( work_dir="./eval_workspace",

Handles all event callbacks and user interactions for the Gradio app.

Top-down fusion block: (optional) residual merge + upsampling + 1x1 contraction

Handles file uploads and processing for the Gradio app.

3DGS parameters, all in world space

Response model for inference API.

Unified inference service class

Model backend service with persistent model loading.

Handles model inference and data processing for Depth Anything 3.

Output processor for converting model outputs to Prediction objects. Handles tensor-to-numpy conversion, batch dimension removal, and cr

Stable PCA→RGB for video features shaped (T, H, W, D) or a single frame (H, W, D). - Global mean/V3 reference for stable colors - Per-fra

Generates and caches 2D spatial positions for patches in a grid. This class efficiently manages the generation of spatial coordinates for patches

2D Rotary Position Embedding implementation. This module applies rotary position embeddings to input tokens based on their 2D spatial positio

Task status model.

Handles UI component creation and layout for the Gradio app.

This is the main model for Visual Place Recognition we use Pytorch Lightning for modularity purposes. Args: pl (_type_): _description

Handles visualization updates and navigation for the Gradio app.

CameraHead predicts camera parameters from token representations using iterative refinement. It applies a series of transformer blocks (the

COLMAP data handler

ANSI escape codes for terminal colors.

DPT for dense prediction (main head + optional sky head, sky always 1 channel). Returns: - Main head: * If output_dim>1: { hea

DTU Benchmark dataset wrapper for DepthAnything3 evaluation. Supports: - Camera pose estimation evaluation (AUC metrics) - 3

DTU-64 Dataset wrapper for DepthAnything3 POSE EVALUATION ONLY. This dataset is a subset of DTU with 64 images per scene. It is specific

Base class for all benchmark datasets. Subclasses must implement: - SCENES: List of scene identifiers - data_root: Path to d

Depth Anything 3 network for depth estimation and camera pose estimation. This network consists of: - Backbone: DinoV2 feature extractor

Dual-head DPT for dense prediction with an always-on auxiliary head. Architectural notes: - Sky/object branches are removed. - `

ETH3D Benchmark dataset wrapper for DepthAnything3 evaluation. Supports: - Camera pose estimation evaluation (AUC metrics) -

HiRoom Dataset wrapper for DepthAnything3 evaluation. Supports: - Camera pose estimation evaluation (AUC metrics) - 3D recon

Single image handler

Image directory handler

Request model for inference API.

Base input handler class

Nested Depth Anything 3 network with metric scaling capabilities. This network combines two DepthAnything3Net branches: - Main branch: S

2D image to patch embedding: (B,C,H,W) -> (B,N,D) Args: img_size: Image size. patch_size: Patch token size. in_chans

ScanNet++ Benchmark dataset wrapper for DepthAnything3 evaluation. Supports: - Camera pose estimation evaluation (AUC metrics)

7Scenes Benchmark dataset wrapper for DepthAnything3 evaluation. Supports: - Camera pose estimation evaluation (AUC metrics)

Video handler