nurtayev-dMIT
mikasa/seasondish_stock224_h264_crf23
MIKASA recordings of MikasaSeasonDish-v0 task with 1000 episodes and 486027 frames at 10 Hz, featuring joint position delta actions and proprioceptive state observations.
Downloads23
Episodes1000
Why This Matters for Physical AI
This dataset provides high-quality demonstration data for training embodied AI agents on complex manipulation tasks with memory requirements, supporting the development of policies that can handle partial observability in robotic control.
Technical Profile
- Modalities
- proprioceptionrgb
- Robot Embodiments
- Mikasa
- Action Space
- joint_delta_positions
- Environment
- simulation
- Task Types
- manipulation
- Episodes
- 1000
- Data Format
- LeRobot
- Annotation Types
- action_labelsreward_labels
- License
- MIT
Access
Need custom proprioception data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack