Skip to content

DATASET.md

Latest commit

 

History

History
134 lines (103 loc) · 3.93 KB

DATASET.md

File metadata and controls

134 lines (103 loc) · 3.93 KB

MimicDroid Dataset Download Instructions

This document provides instructions for downloading the MimicDroid datasets used in the RoboCasa environment.

Dataset Overview

The MimicDroid dataset consists of two main components:

  1. Task Demos Dataset: Contains demonstration data for few-shot learning (~250MB)
  2. Play Data: Contains the full dataset with all task demonstrations (~8GB)

Download Options

Option 1: Using Hugging Face CLI (Recommended)

Use the Hugging Face CLI to download the datasets:

# Install huggingface_hub if not already installed
pip install huggingface_hub

# Download the complete dataset
huggingface-cli download Rutav/MimicDroidDataset --repo-type dataset --local-dir ./MimicDroidDataset

This will:

  • Download the complete MimicDroid dataset from Hugging Face
  • Store it in the specified local directory
  • Handle authentication and progress indicators automatically
  • Allow you to resume interrupted downloads

Option 2: Using Python Script

Use the provided Python script to download the datasets:

python robocasa/scripts/download_mimicdroid_dataset.py

This script will:

  • Download the complete MimicDroid dataset
  • Handle extraction and organization automatically
  • Provide progress indicators during download
  • Allow you to choose which components to download

Option 3: Manual Download via Hugging Face Web Interface

If you prefer to download manually, you can access the dataset directly:

MimicDroid Dataset

To download via web interface:

  1. Visit the repository URL above
  2. Click "Files and versions" tab
  3. Download individual files or use the "Download repository" button

File Structure

After downloading, the datasets will be organized as follows:

PlayDataset

The main dataset containing comprehensive robotic manipulation demonstrations:

MimicDroidDataset/
└── training/                               # Training data and metadata
    ├── 003/                                # Training episode directories
    │   ├── demo_im128_notp.hdf5
    │   └── demo.hdf5
    ├── 004/
    ├── 005/
    ├── 006/
    ├── 007/
    ├── 008/
    ├── 009/
    ├── 010/
    ├── 011/
    ├── 012/
    ├── 013/
    ├── 015/
    ├── 017/
    ├── 018/
    ├── 019/
    ├── 020/
  TaskDemos/
  ├── CloseLeftCabinetDoor/
  │   └── 003/
  │       ├── demo_im128_notp.hdf5
  │       └── demo.hdf5
  ├── CloseLeftCabinetDoorL2/
  ├── CloseLeftCabinetDoorL3/
  ├── CloseRightCabinetDoorL2/
  ├── PnPSinkToCabinet/
  ├── PnPSinkToCabinetL2/
  ├── PnPSinkToMicrowaveTopL3/
  ├── PnPSinkToRightCounterPlate/
  ├── PnPSinkToRightCounterPlateL2/
  ├── PnPSinkToRightCounterPlateL3/
  ├── TurnOnFaucet/
  └── TurnOnFaucetL3/

Requirements

  • Python 3.7+
  • huggingface_hub package (pip install huggingface_hub)
  • Sufficient disk space (at least 10GB recommended)
  • Internet connection for download

Authentication

If the dataset requires authentication:

  1. Create a Hugging Face account at https://huggingface.co
  2. Generate an access token in your account settings
  3. Login using: huggingface-cli login
  4. Enter your access token when prompted

Troubleshooting

If you encounter issues during download:

  1. Ensure you have sufficient disk space
  2. Check your internet connection
  3. Verify you have the huggingface_hub package installed
  4. Try logging in again with huggingface-cli login
  5. Use --resume-download flag to resume interrupted downloads
  6. Check the dataset repository for any access restrictions

Support

For issues related to dataset download or usage, please create an issue in the repository.