VidMap (ECCV 2026) Open-Sourced: Combines SLAM and SfM for Accurate 3D Reconstruction in Long Videos
Scobleizer · x · 2026-08-08
VidMap is a research project accepted to ECCV 2026, with its code and paper now publicly available.
Core Methodology
- Integrates the temporal structure of SLAM into global Structure-from-Motion (SfM).
- Combines temporal tracks, loop closures, metric depth, and global optimization to estimate camera poses, intrinsics, and a sparse 3D map.
- Aims to achieve accurate 3D reconstruction for long, unconstrained videos.
Open Source Details
- Provides a Python package vidmap; requires building COLMAP 4.1 and PyCOLMAP from source.
- Depends on PyTorch and xFormers, tested on Linux x86-64 with NVIDIA GPUs.
More from Research
- PrimeIntellect Launches Multi-Agent Reinforcement Learning Training Stack — shi_weiyan · 2026-08-08
- FactorJEPA: A New World Model for Crowded Urban Environments — Kapil Wanaskar · 2026-08-08
- The Enduring Value of Data Hinges on the Future Cost of Verification and Generation — oyhsu · 2026-08-08
- HKU's Hengshuang Zhao Named MIT TR35 China for Work in Embodied AI and 3D Vision — YiMaTweets · 2026-08-08
- Genomic Intelligence to Demo DNA Models and Agent Integrations in Upcoming Webinar — julia_kiseleva · 2026-08-08
- AI-Generated Patches Fail Half the Time, Study of 6,000+ Patches Finds — WeldPond · 2026-08-08