Axis Robotics has introduced AXIS, a novel browser-based data engine designed to accelerate robot manipulation research. This system allows for continuous data collection through a web interface, with backend GPUs handling computationally intensive tasks like training and rendering. The dataset, currently available on Hugging Face for non-commercial academic use, features over 50,000 trajectories across 207 tasks, generated procedurally to ensure variety and scalability. Initial results show that models trained with AXIS data significantly outperform those trained on static datasets, demonstrating the effectiveness of this dynamic data collection approach. AI
IMPACT This system could accelerate robot learning by enabling continuous, community-driven data collection and improving model performance on manipulation tasks.
RANK_REASON Release of a new dataset and associated tools for robot manipulation research. [lever_c_demoted from research: ic=1 ai=1.0]
- A100
- AXIS
- Axis Robotics
- F ranka Research 3
- Gemma-2B
- Gemma-300M
- Georgia Tech
- Hugging Face
- IsaacSim
- LIBERO-Plus
- MuJoCo
- Nanyang Technological University
- Openpicus
- PaliGemma
- React
- RTX 4090
- TaskGenies
- University of California, Berkeley
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →