You need to agree to share your contact information to access this dataset
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
📋 M3 Multimodal Human-Robot Collaboration Dataset - Access Request
Thank you for your interest in the M³-HRC dataset. To ensure responsible use and track dataset usage, please provide the following information:
By requesting access, you agree to:
- Use this dataset for research purposes only
- Not redistribute the dataset or any part of it
- Cite the dataset in any publications resulting from its use
- Adhere to the CC BY-NC 4.0 license terms
Your information will be used solely for access management and compliance tracking.
Log in or Sign Up to review the conditions and access this dataset content.
M³-HRC: Multimodal Human-Robot Collaboration Dataset
M³-HRC is a multimodal dataset designed for analyzing Human-Robot Collaboration (HRC) in industrial assembly environments. It extends the MIAM dataset with fine-grained action primitives annotations, enabling unified analysis across multiple behavioral scales.
This dataset is introduced in the paper:
**"The Action-Engagement-Collaboration Triad: A Multimodal Analytical Framework for Human-Robot Collaboration"**
Key Features
🔹 Multimodal Data
- Third-person view videos from camera-1 (C1)
- Top-view RGB and depth streams (C2)
- IMU sensor data (separately for Left and Right Hand)
🔹 Fine-Grained Action Annotations
- 280 segmented action primitives
- 17 action categories
Dataset Structure
Dataset Structure 2024-9-20_11-40-54-563/ # Each directory as directory and file stracture │ ├── c1_clips/ # Third-person videos (primitive segmented) ├── c2_rgb_clips/ # Top-view RGB videos ├── c2_depth_clips/ # Depth streams ├── lh_imu_data.csv # Left Hand IMU sensor CSV file ├── rh_imu_data.csv # Right Hand IMU sensor CSV file └── c1_action_annotations.csv # Primitive Action Annotations
Map Operator ID with Raw Directory Names
| Operator ID | Raw Directory Name |
|---|---|
| ID-1 | 2024-9-21_16-55-29-534 , 2024-9-26_18-46-48-169 |
| ID-2 | 2024-9-21_18-02-33-428 |
| ID-3 | 2024-9-25_16-51-01-958 , 2024-9-25_16-58-02-103 |
| ID-4 | 2024-9-27_18-00-07-834 , 2024-9-20_11-40-54-563 , 2024-9-20_12-37-54-780 , 2024-9-20_18-26-57-871 |
| ID-5 | 2024-9-25_18-03-19-106 |
| ID-6 | 2024-12-02_16-50-19-914 , 2024-12-02_17-05-00-612 |
| ID-7 | 2024-9-26_17-59-47-126 |
| ID-8 | 2024-10-05_17-23-20-277 |
🔄 Data Description
- Subjects: 8 operators
- Domain: Industrial assembly tasks
- Annotation Tool: VGG Video Annotator (VIA)
- Inter-Annotator Agreement: κ = 0.85
All modalities are temporally synchronized, referenced by Camera-1 hardware time stamp.
📊 M³ Framework (MICRO-MESO-MACRO)
The dataset supports multi-scale analysis using:
MICRO Layer
- Action entropy
- Diversity ratio
MESO Layer
- Engagement stability
- High engagement ratio
MACRO Layer
- Collaboration score
- Interaction efficiency
📋 Usage Instructions
Step 1: Request Access
This dataset is gated. You must request access and receive approval before downloading.
To request access:
- Click the "Request Access" button above
- Fill out the access request form with your details
- Wait for approval (you'll receive email notification)
- Once approved, you can download the dataset files
Step 2: Download and Extract
After approval, download the required zip files and extract:
unzip filename.zip
📁 Dataset Files The dataset is organized into the following zip archives (by operator sessions):
- 2024-9-20_11-40-54-563.zip (728 MB)
- 2024-9-20_12-37-54-780.zip (346 MB)
- 2024-9-20_18-26-57-871.zip (1.2 GB)
- 2024-9-21_16-55-29-534.zip (2.1 GB)
- 2024-9-21_18-02-33-428.zip (1.7 GB)
- 2024-9-25_16-51-01-958.zip (503.9 MB)
- 2024-9-25_16-58-02-103.zip (203.5 MB)
- 2024-9-25_18-03-19-106.zip (125.0 MB)
- 2024-9-26_17-59-47-126.zip (2.5 GB)
- 2024-9-26_18-46-48-169.zip (1.3 GB)
- 2024-9-27_18-00-07-834.zip (799.6 MB)
- 2024-10-05_17-23-20-277.zip (1.2 GB)
- 2024-12-02_16-50-19-914.zip (1.1 GB)
- 2024-12-02_17-05-00-612.zip (348.6 MB)
Each archive contains the full multimodal data for a specific operator session.
License
This dataset is released under: Creative Commons Attribution-NonCommercial 4.0 (CC BY-NC 4.0) Research use allowed Attribution required Commercial use prohibited Usage Policy Dataset access requires approval For research purposes only Redistribution is not allowed Proper citation is mandatory
Citation
If you use this dataset, please cite:
@inproceedings{arvind2026action,
title={The Action-Engagement-Collaboration Triad: A Multimodal Analytical Framework for Human-Robot Collaboration},
author={Arvind and Mehta, Naval Kishore and Kumar, Himanshu and Saurav, Sumeet and Singh, Sanjay},
booktitle={Proceedings of the 21st ACM/IEEE International Conference on Human-Robot Interaction},
pages={1293--1297},
year={2026}
}
Related Resources Code: https://github.com/arvindsihag/m3_analyzer
Acknowledgements Developed at: CSIR-Central Electronics Engineering Research Institute (CEERI), Pilani Academy of Scientific and Innovative Research (AcSIR), India
Contact For queries, collaborations, or access requests: Arvind Sihag Email: arvind.ceeri24a@acsir.res.in
Notes This dataset enables cross-layer analysis of human behavior in collaborative environments and is intended to support research in Human-Robot Interaction (HRI) Industrial AI Multimodal Learning Behavioral Analytics
- Downloads last month
- 5