SHANGHAI--(BUSINESS WIRE)--Robbyant, an embodied AI company within Ant Group, today announced the launch of LingBot-Depth 2.0, a next-generation spatial perception model, alongside its foundational visual model, LingBot-Vision. This release marks a significant leap in robotic spatial perception, empowering robots to accurately understand and navigate the physical world.

From "Seeing" to "Seeing Accurately"

Building upon the success of LingBot-Depth, which pioneered the Masked Depth Modeling (MDM) technique to resolve depth sensing challenges for transparent and reflective surfaces, LingBot-Depth 2.0 represents a massive scale-up in training data and performance.

Trained on 150 million samples, the new model achieves top rankings in 12 out of 16 depth completion benchmarks. Most notably, in the most demanding indoor scenarios with massive depth loss, LingBot-Depth 2.0 halves the depth error compared to its predecessor, reducing RMSE from 0.132 to 0.062. Furthermore, it demonstrates particularly outstanding performance in traditional failure cases for conventional depth cameras, such as glass, mirrors, and transparent objects.

This breakthrough is driven by LingBot-Vision, a novel visual foundation model and the first in the industry to use "boundary structure" as a pre-training objective. With sub-pixel-level boundary localization and spatial structure understanding, LingBot-Vision provides the precise visual representations required for robust spatial perception.

Remarkably, the pre-training corpus for LingBot-Vision consists of only 160 million images, which is an order of magnitude smaller than that of DINOv3. Moreover, it brings highly stable object boundary determination capability, enabling continuous tracking of object boundaries within videos.

Beyond supporting LingBot-Depth 2.0, LingBot-Vision is a versatile foundation model capable of diverse downstream tasks.

Collaboration with Orbbec on Applications

For commercial applications, LingBot-Depth 2.0 has been professionally certified by the Depth Vision Laboratory of Orbbec, a leading provider of robotics and AI vision. Real-world testing using chip-level depth data from Orbbec’s Gemini 330 series stereo 3D cameras demonstrates significant improvements in edge clarity, object contour integrity, small object recognition, long-range depth estimation, and robustness in complex lighting and material conditions.

Moreover, within the product matrix of Orbbec's newly released Robot-Free Data Collection Hardware Platform, the RGB-D EGO device will integrate a customized LingBot-Depth model optimized for data collection. Moving forward, it will further integrate an advanced commercial version to continuously optimize missing depth completion, object edges, and spatial structures, delivering a precise, stable, and highly usable real-world data foundation for embodied AI training.

In collaboration with Robbyant, Orbbec will launch new products integrating LingBot-Depth 2.0:

• SDK Product: Enabling edge-side deployment for robotics customers, this SDK will significantly enhance depth perception for systems utilizing Gemini 330 series cameras.
• Integrated Camera: Planned for release by year-end, this all-in-one solution will deliver "3D Camera + Spatial Perception" as a unified package.

Additionally, Robbyant has open-sourced the model weights of LingBot-Vision. The company remains committed to building the robotic vision foundation together with industry partners, resolving the critical bottlenecks of "seeing, seeing accurately, and seeing stably" in the real physical world, and accelerating the commercial deployment of embodied intelligence.

For more information about LingBot-Depth 2.0 and LingBot-Vision, please visit:
GitHub: https://github.com/Robbyant/lingbot-vision
Hugging Face: https://huggingface.co/collections/robbyant/lingbot-vision

About Robbyant

Robbyant is an embodied intelligence company within Ant Group, dedicated to advancing embodied intelligence through cutting-edge software and hardware technologies. Robbyant independently develops foundational large models for embodied AI and actively explores next-generation intelligent devices, aiming to create robotic companions and caregivers that truly understand and enhance people’s everyday lives and deliver reliable intelligent services across key use cases, such as elderly care, medical assistance, and household tasks.

To learn more about Robbyant, please visit: www.robbyant.com