tum-vision/lsd_slam

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

2724 stars · 1228 forks observed · 2026-08-28

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

LSD-SLAM is a real-time monocular SLAM system that uses direct (featureless) image alignment to build large-scale, semi-dense 3D maps from a single camera. It is a C++ research codebase built around ROS for input/output, from TUM's computer vision group.

Use cases

• real-time monocular SLAM from a single camera stream
• build semi-dense 3D point cloud maps from video
• visual odometry without feature extraction
• reconstruct indoor environments from a rosbag
• research baseline for direct SLAM methods

When to choose

• you need real-time semi-dense monocular SLAM on a laptop
• your scene lacks distinctive keypoints or textures suitable for feature-based SLAM
• you are reproducing or comparing against the LSD-SLAM papers
• your pipeline already uses ROS for camera input and visualization

When to avoid

• you need a maintained, easily buildable system - it targets old Ubuntu 12.04/14.04 and ROS fuerte/indigo
• you want stereo, RGB-D, or multi-camera SLAM - it is monocular only
• you need dense maps rather than semi-dense ones
• you cannot use ROS, since building without ROS is unsupported

Facets

library · maturity maintenance

computer-vision graphics simulation robotics computer-vision autonomous-vehicles cpp slam monocular-slam visual-odometry direct-method semi-dense-mapping ros 3d-reconstruction real-time linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="tum-vision/lsd_slam")