In April 2023, DAIMON Robotics, a Hong Kong-based company, launched Daimon-Infinity, touted as the world's largest omni-modal robotic dataset for physical AI. This extensive dataset, which includes high-resolution tactile sensing data from over 80 real-world scenarios and 2,000 human skills, aims to enhance robot manipulation capabilities across various tasks, from household chores to industrial assembly lines. The initiative is backed by collaborations with prominent partners, including Google DeepMind, Northwestern University, and the National University of Singapore.
Prof. Michael Yu Wang, co-founder and chief scientist of DAIMON, emphasized the importance of tactile feedback in improving robotic dexterity, advocating for a shift from the traditional Vision-Language-Action (VLA) model to a more integrated Vision-Tactile-Language-Action (VTLA) framework. This transition is crucial for enabling robots to perform complex manipulation tasks effectively, especially in environments where visual data alone is insufficient.
Recognizing a significant data gap in the robotics industry, DAIMON has committed to open-sourcing 10,000 hours of its dataset to support broader research and development efforts. The company aims to accelerate the deployment of embodied AI by providing high-quality tactile data, which is essential for training robots to interact with their surroundings more naturally and effectively. As the robotics landscape evolves, DAIMON's innovative approach positions it as a key player in advancing the capabilities of humanoid robots in real-world applications.
Leave a comment