# Junyi42/monst3r

Official Implementation of paper "MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motion"

Repository: https://github.com/Junyi42/monst3r
Canonical: https://ross.abutalabs.com/products/monst3r
Homepage: https://monst3r-project.github.io/
Language: Python
License: NOASSERTION
License Family: other
Last push: 2025-06-16T21:01:48+00:00

## Health v2 (maintenance only)
Score: 36/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 27, release rhythm 40, longevity 49
- inputs: {"age_days": 695, "days_push": 443, "days_rel": 672, "gap_med": 9, "n_releases_24m": 2}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1386, forks 85 (observed 2026-08-28T04:04:35.039489+00:00)

## What it is
MonST3R is the official PyTorch implementation of an ICLR 2025 paper that estimates per-timestep geometry (pointmaps) from dynamic videos in a feed-forward manner. It produces time-varying dynamic point clouds with per-frame camera poses and intrinsics, enabling downstream tasks like video depth estimation and dynamic/static scene segmentation.

## Use cases
- estimate 3d geometry from dynamic videos with moving objects
- recover per-frame camera poses and intrinsics from video
- generate time-varying dynamic point clouds from footage
- estimate video depth for scenes with motion
- segment dynamic versus static parts of a scene
- visualize 4d reconstructions of dynamic scenes interactively
- run real-time feed-forward 3d reconstruction from video

## When to choose
- you need geometry estimation from videos containing moving or deforming objects
- you want a feed-forward alternative to multi-stage depth-and-flow pipelines
- you need camera pose and depth outputs from monocular video for research
- you want to reproduce or build on the MonST3R paper's results

## When to avoid
- you only need static-scene 3D reconstruction, where the original DUSt3R may suffice
- you need a production-ready application with a polished UI rather than research code
- you lack a GPU or cannot set up CUDA/conda environments
- you need a permissive license for commercial use, since the license is non-standard

## Facets
- artifact type: library
- maturity: active
- function: computer-vision, machine-learning, deep-learning, image-processing, data-visualization
- domain: computer-vision, machine-learning, deep-learning, artificial-intelligence, graphics
- platform: python, cross-platform
- tags: 3d-reconstruction, point-cloud, camera-pose-estimation, depth-estimation, dynamic-scenes, video, dust3r, research-code, 4d-visualization, iclr-2025, linux, gpu

## Member repositories
- Junyi42/monst3r (main) score 36

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:35.039489+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:39:52.798763+00:00, confidence not recorded.
  - readme: https://github.com/Junyi42/monst3r (fetched 2026-08-28T04:04:35.039489+00:00, sha 257fe18af422)
  - homepage: https://monst3r-project.github.io/ (fetched 2026-08-29T11:55:20.393263+00:00, sha 6ed1550b3880)
- Data as of 2026-08-30T08:39:29.467469+00:00.