facebookresearch / facebookresearch/pytorch3d

[feature request] Oriented bounding box computation

Open
#807 1 comment 1 reaction 0 assignees View on GitHub
enhancement
Dominant language
Python
Stars
10k
Forks
1.5k
PR merge metrics
No merged PRs in 30d

Description

This is useful in many contexts (e.g. to compare visually poses of different 3D reconstruction methods).

Ideally, it would directly support different representations: meshes, 3d point clouds.

This exists in CGAL: https://doc.cgal.org/latest/Optimal_bounding_box/index.html, but CGAL is quite hard to install

I also found this functionality in Open3D: http://www.open3d.org/docs/latest/python_api/open3d.geometry.OrientedBoundingBox.html and in trimesh: https://github.com/mikedh/trimesh/blob/3bbe6b368c723f171b9d1aa7c58cfbaf77279546/trimesh/bounds.py#L108

Contributor guide

Open the contributing guide

Research direction

The issue names no repository files or tests. Start by comparing the linked CGAL, Open3D, and trimesh implementations and tracing how PyTorch3D represents meshes and 3D point clouds. Done means an agreed API and implementation plan for oriented bounding boxes covering the requested representations without requiring CGAL.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
computer-graphics
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.