Showing 28 datasets
Sort: Newest ▾
High-fidelity, time-synchronized ego-centric and tactile datasets powering the next generation of AI. Built for pretraining and industrial-grade VLA model deployment.
We know the ceiling of a foundation model is set by its data. Every dataset passes a rigorous 5-step QC process.
Tactile, vision and IMU streams synchronized to sub-millisecond precision.
Complete annotations for vision-language-action models, including semantic object tracking.
Sensor noise and gravity-bias filtering applied to raw IMU and force feedback data.
Automatic face blurring and masking of sensitive personal information across all visual modalities.
