droid_1.0.1
Verified95,600 Franka episodes and 27M frames recorded at 15 FPS.
What is droid_1.0.1?
droid_1.0.1 is a robotics dataset collected with a Franka robot arm. It comprises 95,600 episodes totaling 27,612,581 frames at 15 FPS, stored in chunked parquet files with associated videos.
It supports training and evaluation of imitation learning models and robotic control policies on manipulation tasks.
What you can build with droid_1.0.1
Train imitation learning policies
Use the 27M+ frames of Franka robot trajectories to train end-to-end policies that map observations to actions via behavioral cloning.
Benchmark offline RL algorithms
Evaluate reinforcement learning methods on the 95,600 episodes split into train set, measuring success on manipulation tasks with the provided video and state data.
Develop video-conditioned controllers
Leverage the 286,800 videos at 15 fps to build models that predict actions from visual input for Franka arm control.
Load droid_1.0.1
from datasets import load_dataset
ds = load_dataset("cadene/droid_1.0.1")- 1Install the Hugging Face datasets library with pip install datasets
- 2Load the dataset via load_dataset('cadene/droid_1.0.1')
- 3Read meta/info.json to inspect robot_type, fps, and splits
- 4Access episode data through the parquet paths in data/chunk-*/episode_*.parquet
- 5Process videos using the video_path template for LeRobot-compatible workflows
droid_1.0.1: pros & cons
Pros
- +Extremely large scale with 95k episodes and 27M frames
- +Includes synchronized video recordings (286k total)
- +Chunked parquet storage for efficient partial loading
- +Direct compatibility with LeRobot codebase
Cons
- –Requires massive storage due to total size
- –Robot type fixed to Franka only
- –total_tasks field is zero, limiting task-level filtering
Frequently asked questions
A large-scale robotics dataset of Franka arm trajectories created with LeRobot and hosted on Hugging Face.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…
Similar datasets
Other ai & machine learning options worth comparing.