LingBot-Depth: Masked Depth Modeling for Metric-Accurate 3D Perception in Robotics
tom_doerr · x · 2026-08-15
The open-source project LingBot-Depth transforms incomplete and noisy depth sensor data into high-quality, metric-accurate 3D measurements. By jointly aligning RGB appearance and depth geometry in a unified latent space, it serves as a spatial perception foundation for robot learning and 3D vision. Applications include depth completion, scene reconstruction, 4D point tracking, and dexterous manipulation. The project provides a technical report, code, data, and model weights.
More from Embodied
- Open-source humanoid achieves zero-shot sim-to-real, works on first try with new engine — freelerobot · 2026-08-15
- Blind Skateboarder: PhD robotics project works without vision or markers — ___Mufasaa · 2026-08-15
- Unitree Prepares for Shanghai IPO as Americans Bid Millions Despite US Ban — 96Stats · 2026-08-15
- Jones Foods AI achieves fully autonomous protein processing — repligate · 2026-08-15
- AI sunglasses can 'erase' people from real-world footage to protect privacy — armano · 2026-08-15
- Real-time BCI meditation system adapts video to brain activity — uisato · 2026-08-15