geohot/twitchslam

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1003 stars · 212 forks observed · 2026-09-03

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

A toy monocular SLAM (Simultaneous Localization and Mapping) implementation written in Python during livestreams. It extracts features from video frames, builds a 3D point map, and estimates camera pose using OpenCV, SDL2, Pangolin, and g2o bindings.

Use cases

• learn how monocular SLAM works from readable code
• run visual odometry on a video file
• experiment with feature extraction and pose estimation
• test SLAM against KITTI or TUM freiburg datasets
• prototype a simple 3D map from camera video

When to choose

• you want a small, hackable SLAM codebase to study or modify
• you're learning computer vision concepts like essential matrices and bundle adjustment
• you need a starting point for a custom visual SLAM experiment

When to avoid

• you need a production-grade or accurate SLAM system
• you require robustness, automated tests, or long-term maintenance
• you need stereo, RGB-D, or real-time performance guarantees

Facets

library · maturity experimental

computer-vision image-processing graphics simulation computer-vision robotics education python slam monocular-slam livestream-project feature-tracking pose-estimation toy-implementation algorithms linux desktop

Member repositories

For agents

markdown · JSON · MCP: product_card(name="geohot/twitchslam")