Overview
LingBot-Depth is a high-precision spatial perception model designed to enhance robots' depth sensing and 3D environmental understanding capabilities in complex real-world environments. Developed by Robbyant, an embodied AI company within Ant Group, it addresses challenges like missing depth information on transparent or reflective surfaces. It uses Masked Depth Modeling (MDM) to infer and reconstruct missing depth regions from RGB image features, producing denser and more accurate 3D maps. Benchmarked against major models like PromptDA and PriorDA, it reduces relative error by over 70% in indoor scenes and RMSE by 47% on sparse Structure-from-Motion tasks.
Key facts
- Maturity
- prototype
Detailed specifications
Other4
- REL Reduction
- 70%
- RMSE Reduction
- 47%
- Company Country
- CN
- Additional Information
- - Masked Depth Modeling: Self-supervised pre-training via depth reconstruction. - Cross-Modal Attention: Joint RGB-Depth alignment in unified latent space. - Metric-Scale Preservation: Maintains real-world measurements for downstream tasks. - Training Data: Includes 2M real-world and 1M simulated RGB-D samples. - Hardware Setup: Scalable RGB-D capture system with Intel RealSense, Orbbec Gemini, and Azure Kinect.
Reviews for LingBot-Depth
Loading reviews…