whatisgithub

What is boxer?

facebookresearch/boxer — explained in plain English

Analysis updated 2026-07-17 · repo last pushed 2026-06-05

★ 580PythonAudience · developerComplexity · 3/5MaintainedSetup · moderate

In one sentence

Boxer turns 2D object detections in photos into 3D bounding boxes for indoor scenes, giving position, orientation, and size for use in AR, robotics, and spatial computing apps.

Mindmap

mindmap
  root((Boxer))
    What it does
      2D to 3D boxes
      Position and size
      Indoor scenes
      Video tracking
    Tech stack
      Python
      OWL detector
      BoxerNet
      CUDA and MPS
    Use cases
      AR anchoring
      Robot navigation
      Scene research
      Batch processing
    Audience
      AR developers
      Roboticists
      Researchers

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

What do people build with it?

USE CASE 1

Anchor digital AR objects to real furniture detected in 3D.

USE CASE 2

Give a robot an understanding of where furniture sits in a room.

USE CASE 3

Build a 3D indoor scene-understanding pipeline for research.

USE CASE 4

Batch-process photos or video into consistent 3D bounding boxes.

What is it built with?

PythonOWL detectorBoxerNetCUDAMPS

How does it compare?

facebookresearch/boxertencent-hunyuan/unirlfacebookresearch/vicreg
Stars580584574
LanguagePythonPythonPython
Last pushed2026-06-05—2023-07-06
MaintenanceMaintained—Dormant
Setup difficultymoderate—hard
Complexity3/5—4/5
Audiencedeveloperresearcherresearcher

Figures from each repo's GitHub metadata at analysis time.

How do you get it running?

Difficulty · moderate Time to first run · 30min

Inference-only with pretrained models, needs camera intrinsics and ideally a depth map, plus a dataset loader for your image format.

Copy-paste prompts

Prompt 1
Show me how to run Boxer's OWL detector to find all chairs in a photo and lift them into 3D bounding boxes.
Prompt 2
Help me write a data loader for Boxer so I can run it on my own RGB-D camera images.
Prompt 3
Explain how to supply camera intrinsics and a depth map to Boxer so it outputs accurate 3D box positions.
Prompt 4
Walk me through using Boxer's multi-frame fusion to track a 3D box across a video sequence.
Prompt 5
Help me set up Boxer to run on my Mac using MPS acceleration.

Frequently asked questions

What is boxer?

Boxer turns 2D object detections in photos into 3D bounding boxes for indoor scenes, giving position, orientation, and size for use in AR, robotics, and spatial computing apps.

What language is boxer written in?

Mainly Python. The stack also includes Python, OWL detector, BoxerNet.

Is boxer actively maintained?

Maintained — commit in last 6 months (last push 2026-06-05).

How hard is boxer to set up?

Setup difficulty is rated moderate, with roughly 30min to a first successful run.

Who is boxer for?

Mainly developer.

Open on GitHub → Ask about another repo

This repo across BitVibe Labs

Verify against the repo before relying on details.

What are similar repos?