Sign in to view source links and access this dataset
Description
RoboCasa365 Target Composite Seen contains 8,077 episodes of human teleoperation data for 16 composite tasks, recorded across 10 held-out target kitchens. The dataset includes 6,002,265 frames, representing approximately 83 hours of video at 20 frames per second. It was created by ember-lab-berkeley and last updated on Hugging Face in May 2026.
Use Cases
Train imitation learning models based on human teleoperation demonstrations.
Benchmark robotic policy performance on composite kitchen tasks.
Develop and evaluate multi-task robotic manipulation systems.
Analyze human behavior patterns for complex, multi-step object manipulation.
Strengths
8,077 episodes provide a substantial number of demonstrations.
6,002,265 frames offer detailed temporal coverage for video-based learning.
Tasks are represented with 500 demonstrations per task, suggesting consistent coverage.
Data is recorded in 10 distinct target kitchens, which may indicate environmental diversity.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is unknown, which may limit suitability assessment.
Freshness should be verified as the last update was in 2026.
Provenance
Source
ember-lab-berkeley
Collection Method
Human teleoperation recordings.
Freshness
Last updated 2026-05-11 02:31:21
License is unknown; users should verify terms before use.