Ant Group's Lingbo Tech Launches Upgraded Spatial Perception Model LingBot-Depth 2.0
On July 7, Lingbo Technology, a humanoid intelligence company under Ant Group, released its spatial perception model, LingBot-Depth 2.0. This new model has undergone comprehensive upgrades, particularly in areas such as edge clarity, recognition of small objects, depth estimation at long distances, and robustness in complex environments. The training for LingBot-Depth 2.0 was conducted using a dataset of 150 million data points, significantly enhancing its capabilities. In parallel with the model's release, Lingbo also introduced LingBot-Vision, a visual foundation model that complements LingBot-Depth 2.0. This dual release aims to establish a complete capability chain for robots, moving from 'understanding' visual input to 'accurately perceiving' it. The overarching goal is to address fundamental challenges in robot vision, including spatial perception, detailed object recognition, and adaptability to intricate surroundings.
The release of LingBot-Depth 2.0 by Ant Group's Lingbo Technology signifies a strategic advancement in the field of embodied AI and robotics. By enhancing spatial perception capabilities with a large-scale dataset and a complementary vision foundation model, Lingbo aims to tackle critical issues in robotic environmental interaction. This development aligns with the broader trend of equipping robots with more sophisticated sensory processing, crucial for their integration into diverse real-world applications. The focus on robustness and fine-grained recognition suggests an effort to bridge the gap between simulated environments and the complexities of physical spaces, a key hurdle for widespread robotic adoption in the coming decade. The company's approach of building a comprehensive 'understanding' to 'accurate perception' pipeline highlights the interconnectedness of perception and action in intelligent systems.
AI-generated to prompt reflection — not editorial opinion, not advice, not a statement of fact. How this works.