azaracla/community_dataset_v1

--- license: apache-2.0 tags: - robotics - smolvla - community - vlab - so100 - manipulation - lerobot - vision-language-action - embodied-ai task_categories: - robotics language: - en size_categories: - 100K<n<1M pretty_name: Community Dataset v1 (v3.0) --- # Community Dataset v1 (v3.0) A large-scale community-contributed robotics dataset for vision-language-action learning, featuring **119 datasets** from **52 contributors** worldwide. This is a converted and curated version of the original [HuggingFaceVLA/community_dataset_v1](https://huggingface.co/datasets/HuggingFaceVLA/community_dataset_v1), upgraded to LeRobot v3.0 format. This dataset was used to pretrain [SmolVLA](https://huggingface.co/lerobot/smolvla_base). It was filtered using specific criteria including fps, minimum number of episodes, and qualitative assessment of video quality, using the [FilterLeRobotData tool](https://huggingface.co/spaces/Beegbrain/FilterLeRobotData). ## 🌟 Overview This dataset represents a collaborative effort from the robotics and AI community to build comprehensive training data for embodied AI systems. Each contribution contains demonstrations of robotic manipulation tasks with the SO100 arm, recorded using [LeRobot tools](https://github.com/huggingface/lerobot), primarily focused on tabletop scenarios and everyday object interactions. ## 📊 Dataset Statistics | Metric | Value | |--------|-------| | **Total Datasets** | 119 | | **Total Episodes** | 9,528 | | **Total Frames** | 4,489,949 | | **Contributors** | 52 | | **Average FPS** | 30 | | **Average Episodes per Dataset** | 80 | | **Primary Tasks** | Manipulation, Pick & Place, Sorting | | **Robot Types** | SO-100 (various colors) | | **Data Format** | LeRobot v3.0 dataset format | | **Total Size** | ~107 GB | ## 🗂️ Structure The dataset maintains a clear hierarchical structure: ``` community_dataset_v1/ ├── contributor1/ │ ├── dataset_name_1/ │ │ ├── data/ # Parquet files with observations │ │ ├── videos/ # MP4 recordings │ │ └── meta/ # Metadata and info │ └── dataset_name_2/ ├── contributor2/ │ └── dataset_name_3/ └── ... ``` Each dataset follows the LeRobot v3.0 format standard, ensuring compatibility with existing frameworks and easy integration. ## 🚀 Usage **1. Authenticate with Hugging Face** You need to be logged in to access the dataset: ```bash # Login to Hugging Face huggingface-cli login # Or alternatively, set your token as an environment variable # export HF_TOKEN=your_token_here ``` Get your token from [https://huggingface.co/settings/tokens](https://huggingface.co/settings/tokens) ### Download the Dataset ```bash hf download username/community_dataset_v1 \ --repo-type=dataset \ --local-dir /path/local_dir/community_dataset_v1 ``` ### Load Individual Datasets ```python from lerobot.datasets.lerobot_dataset import LeRobotDataset import os # Browse available datasets for contributor in os.listdir("./community_dataset_v1"): contributor_path = f"./community_dataset_v1/{contributor}" if os.path.isdir(contributor_path): for dataset in os.listdir(contributor_path): print(f"📁 {contributor}/{dataset}") # Load a specific dataset (requires authentication) dataset = LeRobotDataset( repo_id="local", root="./community_dataset_v1/contributor_name/dataset_name" ) # Access episodes and observations print(f"Episodes: {len(dataset.episode_indices)}") print(f"Total frames: {len(dataset)}") ``` ### Integration with SmolVLA pretraining framework This dataset is designed for training VLA models. You can download this dataset and use it for Vision Language Action Models training framework, [VLAb](https://github.com/huggingface/VLAb/tree/main): 1. Visit the VLAb repository. 2. Follow the training instructions in the repo 3. Point the training script to this dataset ```python accelerate launch --config_file accelerate_configs/multi_gpu.yaml \ src/lerobot/scripts/train.py \ --policy.type=smolvla2 \ --policy.repo_id=HuggingFaceTB/SmolVLM2-500M-Video-Instruct \ --dataset.repo_id="username/community_dataset_v1/AndrejOrsula/lerobot_double_ball_stacking_random,username/community_dataset_v1/aimihat/so100_tape" \ --dataset.root="local/path/to/datasets" \ --dataset.video_backend=pyav \ --dataset.features_version=2 \ --output_dir="./outputs/training" \ --batch_size=8 \ --steps=200000 \ --wandb.enable=true \ --wandb.project="smolvla2-training" ``` ## 🔧 Dataset Format (v3.0) Each dataset contains: - **`data/`**: Parquet files with timestamped observations - Robot states (joint positions, velocities) - Action sequences - Camera observations (multiple views) - Language instructions - **`videos/`**: Synchronized video recordings - Multiple camera angles - High-resolution capture - Timestamp alignment - **`meta/`**: Metadata and configuration - Dataset info (fps, episode count) - Robot configuration - Task descriptions ### Key Differences from v2.1 - **Unified data files**: Episodes are concatenated into fewer parquet files (improved I/O) - **Restructured metadata**: Episodes and stats stored in Parquet format instead of JSONL - **Improved video organization**: Videos reorganized by camera key for better streaming ## 🎯 Intended Use This dataset is designed for: - **Vision-Language-Action (VLA) model training** - **Robotic manipulation research** - **Imitation learning experiments** - **Multi-task policy development** - **Embodied AI research** ## 🤝 Community Contributions This dataset exists thanks to the generous contributions from researchers, hobbyists, and institutions worldwide. Each dataset represents hours of careful data collection and curation. ### Contributing Guidelines Future contributions should follow: - LeRobot v3.0 dataset format - Consistent naming conventions for the features, camera views etc. - Quality validation checks - Proper task descriptions, describing the actions precisely. Check the [blogpost](https://huggingface.co/blog/lerobot-datasets) for more information ## 🔗 Related Work - [VLAb Framework](https://github.com/huggingface/VLAb) - [SmolVLA model](https://huggingface.co/lerobot/smolvla_base) - [SmolVLA Blogpost](https://huggingface.co/blog/smolvla) - [SmolVLA Paper](https://huggingface.co/papers/2506.01844) - [Docs](https://huggingface.co/docs/lerobot/smolvla) - [How to Build a successful Robotics dataset with Lerobot?](https://huggingface.co/blog/lerobot-datasets) - [Original Community Dataset v1 (v2.1)](https://huggingface.co/datasets/HuggingFaceVLA/community_dataset_v1) --- *Converted and curated with ❤️ by the LeRobot Community*许可证：Apache-2.0 标签： - 机器人学 - SmolVLA - 社区 - VLab - SO100 - 操作任务 - LeRobot - 视觉-语言-动作（Vision-Language-Action） - 具身人工智能（Embodied-AI） 任务类别： - 机器人学 语言： - 英语 规模类别： - 100K<n<1M 展示名称：社区数据集v1（v3.0） --- # 社区数据集v1（v3.0） 这是一款大规模社区贡献式机器人学数据集，用于视觉-语言-动作（Vision-Language-Action）学习，收录了来自全球52位贡献者的**119个数据集**。本数据集是原始 [HuggingFaceVLA/community_dataset_v1](https://huggingface.co/datasets/HuggingFaceVLA/community_dataset_v1) 的转换与整理版本，已升级至LeRobot v3.0格式。 本数据集曾用于预训练 [SmolVLA](https://huggingface.co/lerobot/smolvla_base)。数据集通过 [FilterLeRobotData工具](https://huggingface.co/spaces/Beegbrain/FilterLeRobotData)，基于帧率、最小回合数以及视频质量的人工评估等标准完成筛选。 ## 🌟 概述 本数据集是机器人学与人工智能社区协作的成果，旨在为具身人工智能（Embodied AI）系统构建全面的训练数据。每一份贡献均包含使用 [LeRobot工具](https://github.com/huggingface/lerobot) 录制的SO100机械臂操作任务演示，主要聚焦于桌面场景与日常物品交互。 ## 📊 数据集统计 | 指标 | 数值 | |--------|-------| | **总数据集数** | 119 | | **总回合数** | 9,528 | | **总帧数** | 4,489,949 | | **贡献者数量** | 52 | | **平均帧率** | 30 | | **单数据集平均回合数** | 80 | | **核心任务** | 操作任务、抓取放置、分拣 | | **机器人类型** | SO-100（多种配色） | | **数据格式** | LeRobot v3.0数据集格式 | | **总容量** | 约107 GB | ## 🗂️ 数据结构 本数据集采用清晰的层级结构： community_dataset_v1/ ├── contributor1/ │ ├── dataset_name_1/ │ │ ├── data/ # 存储观测数据的Parquet文件 │ │ ├── videos/ # MP4格式录制视频 │ │ └── meta/ # 元数据与相关信息 │ └── dataset_name_2/ ├── contributor2/ │ └── dataset_name_3/ └── ... 每个数据集均遵循LeRobot v3.0格式标准，可与现有框架兼容并轻松集成。 ## 🚀 使用方法 **1. 登录Hugging Face** 如需访问本数据集，需完成登录： bash # 登录Hugging Face huggingface-cli login # 或通过设置环境变量的方式 # export HF_TOKEN=your_token_here 可从 [https://huggingface.co/settings/tokens](https://huggingface.co/settings/tokens) 获取您的访问令牌。 ### 下载数据集 bash hf download username/community_dataset_v1 --repo-type=dataset --local-dir /path/local_dir/community_dataset_v1 ### 加载单个数据集 python from lerobot.datasets.lerobot_dataset import LeRobotDataset import os # 浏览可用数据集 for contributor in os.listdir("./community_dataset_v1"): contributor_path = f"./community_dataset_v1/{contributor}" if os.path.isdir(contributor_path): for dataset in os.listdir(contributor_path): print(f"📁 {contributor}/{dataset}") # 加载指定数据集（需完成登录） dataset = LeRobotDataset( repo_id="local", root="./community_dataset_v1/contributor_name/dataset_name" ) # 访问回合与观测数据 print(f"回合数: {len(dataset.episode_indices)}") print(f"总帧数: {len(dataset)}") ### 与SmolVLA预训练框架集成 本数据集专为训练视觉-语言-动作模型设计，可配合 [VLAb](https://github.com/huggingface/VLAb/tree/main) 视觉语言动作模型训练框架使用： 1. 访问VLAb代码仓库 2. 按照仓库内的训练说明完成操作 3. 将训练脚本指向本数据集路径 python accelerate launch --config_file accelerate_configs/multi_gpu.yaml src/lerobot/scripts/train.py --policy.type=smolvla2 --policy.repo_id=HuggingFaceTB/SmolVLM2-500M-Video-Instruct --dataset.repo_id="username/community_dataset_v1/AndrejOrsula/lerobot_double_ball_stacking_random,username/community_dataset_v1/aimihat/so100_tape" --dataset.root="local/path/to/datasets" --dataset.video_backend=pyav --dataset.features_version=2 --output_dir="./outputs/training" --batch_size=8 --steps=200000 --wandb.enable=true --wandb.project="smolvla2-training" ## 🔧 数据集格式（v3.0） 每个数据集包含以下内容： - **`data/`**：带时间戳的观测数据Parquet文件 - 机器人状态（关节位置、速度） - 动作序列 - 多视角相机观测数据 - 语言指令 - **`videos/`**：同步录制的视频 - 多相机角度 - 高分辨率采集 - 时间戳对齐 - **`meta/`**：元数据与配置信息 - 数据集信息（帧率、回合数） - 机器人配置 - 任务描述 ### 与v2.1版本的主要差异 - **统一数据文件**：将多个回合合并为更少的Parquet文件（优化I/O性能） - **重构元数据结构**：回合与统计信息以Parquet格式存储，而非JSONL格式 - **优化视频组织**：按相机键重新组织视频，便于流式加载 ## 🎯 预期用途 本数据集适用于： - **视觉-语言-动作（VLA）模型训练** - **机器人操作任务研究** - **模仿学习实验** - **多任务策略开发** - **具身人工智能研究** ## 🤝 社区贡献 本数据集的诞生离不开全球各地研究者、爱好者与机构的慷慨贡献。每一份数据集都凝聚了数小时的精心数据采集与整理工作。 ### 贡献指南 未来的贡献需遵循以下规范： - 采用LeRobot v3.0数据集格式 - 特征、相机视角等内容采用统一命名约定 - 完成质量验证检查 - 提供准确的任务描述，清晰说明动作内容 可查阅 [博客文章](https://huggingface.co/blog/lerobot-datasets) 获取更多信息。 ## 🔗 相关工作 - [VLAb框架](https://github.com/huggingface/VLAb) - [SmolVLA模型](https://huggingface.co/lerobot/smolvla_base) - [SmolVLA博客文章](https://huggingface.co/blog/smolvla) - [SmolVLA论文](https://huggingface.co/papers/2506.01844) - [官方文档](https://huggingface.co/docs/lerobot/smolvla) - [如何使用LeRobot构建高质量机器人数据集？](https://huggingface.co/blog/lerobot-datasets) - [原始社区数据集v1（v2.1）](https://huggingface.co/datasets/HuggingFaceVLA/community_dataset_v1) --- *由LeRobot社区倾情转换与整理 ❤️*