DGU-HAO: A Dataset with Daily Life Objects for Comprehensive 3D Human Action Analysis (PDF)

This repository holds the codebase, dataset, and models for the work: MMNet: A Model-based Multimodal Network for Human Action Recognition in RGB-D Videos Bruce X.B. Yu, Yan Liu, Xiang Zhang, Sheng-hua Zhong, Keith C.C. Chan, TPAMI 2022 (PDF)

We built the DGU-HAO: A Dataset with Daily Life Objects for Comprehensive 3D Human Action Analysis dataset and validated it using the MMNet: A Model-based Multimodal Network for Human Action Recognition in RGB-D Videos.

Dataset Download

Raw Data & Labeled Data

Data access

(To access the data, please use a VPN to change your location to South Korea and then access the link above.)

Pre-processed Skeleton data

We pre-processed human motion capture data (bvh, json) to the skeleton data with 25 joints.

We have 67,505 samples of pre-processed skeleton data.

Pre-processed skeleton data access

(To access the data, please use a VPN to change your location to South Korea and then access the link above.)

Dataset Structure

Regular expression of the data file name

.
├── raw_data
│   ├── EL
│   │   │── HE
│   │   │   ├── HE01
│   │   │   │   ├── A16
│   │   │   │   │   ├── F
│   │   │   │   │   │   ├── MH01
│   │   │   │   │   │   │   ├── .jpg
│   │   │   │   │   │   │   ├── .mp4
│   │   │   │   │   │   │   ├── .bvh
│   │   │   │   │   │   │   ├── .qtm
│   │   │   │   │   │   │   ├── .jpg
│   │   │   │   │   │   │   ├── .mp4
│   │   │   │   │   │   │   ├── .bvh
│   │   │   │   │   │   │   ├── .qtm
│   │   │   │   │   │   │   └── ...
│   │   │   │   │   │   └── ...
│   │   │   │   │   └── M
│   │   │   │   └── ...
│   │   │   └── ...
│   │   │
│   │   └── OE
│   │       │── ...
│   ├── FN
│   │   │── ...
│   └── 3D_object_modeling
│       │── CH
│       │   │── CH01
│       │   │   │── BaseColor.png
│       │   │   │── Metallic.png
│       │   │   │── Normal.png
│       │   │   │── Roughness.png
│       │   │   └── .fbx
│       │   └── ...
│       └── ...
│ 
└── labeled_data
    ├── EL
    │   │── HE
    │   │   ├── HE01
    │   │   │   ├── A16
    │   │   │   │   ├── F
    │   │   │   │   │   ├── MH01
    │   │   │   │   │   │   ├── .json
    │   │   │   │   │   │   ├── ...
    │   │   │   │   │   │   └── .json
    │   │   │   │   │   └── ...
    │   │   │   │   └── M
    │   │   │   └── ...
    │   │   └── ...
    │   │
    │   └── OE
    │       │── ...
    └── FN
        │── ...

Action Classes (TBU)

Environment Setting

1. Clone the repository

git clone https://github.com/CSID-DGU/NIA-MoCap-1.git
cd NIA-MoCap-1

2. Install the requriements

Prerequisites

• Python3 (>3.5)
• PyTorch depending on user's machine.

pip install -r requirement.txt

cd torchlight; python setup.py install; cd ..

3. Dataset Citation

@article{park2024dgu,
  title={DGU-HAO: A Dataset With Daily Life Objects for Comprehensive 3D Human Action Analysis},
  author={Park, Jiho and Kim, Junghye and Gil, Yujung and Kim, Dongho},
  journal={IEEE Access},
  year={2024},
  publisher={IEEE}