droid_lerobot
VerifiedLarge LeRobot dataset of 92k Franka robot episodes and 27M frames.
What is droid_lerobot?
droid_lerobot is a robotics dataset generated with the LeRobot framework. It provides 92k episodes, 27M frames, and video recordings from a Franka robot across 31k tasks.
It supports machine learning research on robot control, imitation learning, and manipulation tasks using the provided episode and video structure.
What you can build with droid_lerobot
Train imitation learning policies
Use the 92k+ Franka episodes to train visuomotor policies that map RGB video frames to robot actions for manipulation tasks.
Benchmark robot learning algorithms
Evaluate new methods on the 31k tasks with standardized parquet data and video paths for reproducible results.
Develop task-conditioned models
Leverage the task diversity to build models that generalize across hundreds of different robot behaviors recorded at 15 FPS.
Load droid_lerobot
from datasets import load_dataset
ds = load_dataset("IPEC-COMMUNITY/droid_lerobot")- 1pip install datasets
- 2from datasets import load_dataset
- 3dataset = load_dataset('IPEC-COMMUNITY/droid_lerobot')
- 4Access episodes via dataset['train'] and load videos from parquet paths
- 5Use LeRobot v2.0 utilities for frame-level processing
droid_lerobot: pros & cons
Pros
- +Very large scale with 27M frames
- +Diverse tasks and native video support
- +Ready-to-train split in efficient parquet format
- +Direct Franka robot data from LeRobot pipeline
Cons
- –No validation or test split provided
- –Tied to single robot embodiment
- –Large total size requires significant storage
Frequently asked questions
A large-scale Franka robot dataset with 92k episodes and 276k videos in parquet format, created with LeRobot v2.0 for training.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…
Similar datasets
Other ai & machine learning options worth comparing.