task_categories:
- robotics
tags:
- robotics
- humanoid-robot
- manipulation
- in-context-learning
- simulation
- robocasa
MimicDroid: In-Context Learning for Humanoid Robot Manipulation from Human Play Videos
Paper | Project Page | Code
This repository hosts the dataset used in the MimicDroid project. MimicDroid aims to enable humanoid robots to efficiently solve new manipulation tasks from a few video examples. It leverages human play videos—continuous, unlabeled videos of people interacting freely with their environment—as a scalable and diverse training data source for in-context learning (ICL) policies, thereby reducing reliance on labor-intensive teleoperated data.
The project introduces a benchmark built on RoboCasa, a large-scale simulation framework for training generally capable robots to perform everyday tasks.
Dataset Details
The MimicDroid dataset spans 30 objects, 8 kitchen environments, and 8 hours of human play data for training. All the training environments are shown below:
Evaluation is structured into three levels with increasing difficulty and 4 tasks in each level:
| Level | Task Name | Abstract Embodiment | Humanoid Embodiment |
|---|---|---|---|
| L1 (Seen Objects, Seen Environment) | PnPSinkToRightCounterPlate |  |
 |
| PnPSinkToCabinet |  |
 |
|
| TurnOnFaucet |  |
 |
|
| CloseLeftCabinetDoor |  |
 |
|
| L2 (Unseen Objects, Seen Environment) | PnPSinkToRightCounterPlateL2 |  |
 |
| PnPSinkToCabinetL2 |  |
 |
|
| CloseRightCabinetDoorL2 |  |
 |
|
| CloseLeftCabinetDoorL2 |  |
 |
|
| L3 (Unseen Objects, Unseen Environment) | CloseLeftCabinetDoorL3 |  |
 |
| PnPSinkToRightCounterPlateL3 |  |
 |
|
| PnPSinkToMicrowaveTopL3 |  |
 |
|
| TurnOnFaucetL3 |  |
 |
Dataset Download
To download the necessary kitchen assets for the dataset (approximately 5GB), use the following command:
python robocasa/scripts/download_kitchen_assets.py
For more detailed dataset installation and visualization instructions, please refer to the DATASET.md file in the GitHub repository.
Citation
If you find this dataset or research useful for your work, please cite the following:
@article{shah2025mimicdroid,
title={MimicDroid: In-Context Learning for Humanoid Manipulation from Human Play Videos},
author={Shah, Rutav and Liu, Shuijing and Wang, Qi and Jiang, Zhenyu and Kumar, Sateesh and Seo, Mingyo and Mart{\'\i}n-Mart{\'\i}n, Roberto and Zhu, Yuke},
journal={arXiv preprint arXiv:2509.09769},
year={2025}
}
@inproceedings{robocasa2024,
title={RoboCasa: Large-Scale Simulation of Everyday Tasks for Generalist Robots},
author={Soroush Nasiriany and Abhiram Maddukuri and Lance Zhang and Adeet Parikh and Aaron Lo and Abhishek Joshi and Ajay Mandlekar and Yuke Zhu},
booktitle={Robotics: Science and Systems},
year={2024}
}