MonoEye-Dataset

This repository is for the dataset of "MonoEye: Multimodal Human Motion Capture System Using A Single Ultra-Wide Fisheye Camera" (UIST 2020).

Download Dataset [MacOS and Linux]

You need around 600GB of storage to download the dataset contains compressed and decompressed files. Make sure wget is installed:

# on Mac OS
brew install wget

To download the dataset, you need to agree the license in LICENSE.md and set the flag and download path in conf.ig file. Then use the download.py python script:

python ./download.py

Decompress the dataset with following command:

cat ./DOWNLOAD_PATH/train-set.* | tar xvfz -

Structure

Path

The synthetic dataset is structured as follows (train, valid, and test sets.):

synthetic
├── train
│   ├── body
│   │   └── 00000000.png
│   │   └── ...
│   │   └── gt.csv
│   ├── head
│   │   └── 00000000.png
│   │   └── ...
│   │   └── gt.csv
│   ├── camera
│   │   └── 00000000.png
│   │   └── ...
│   │   └── gt.csv
│
├── valid
│   ├── body
│   ├── head
│   ├── camera
│
├── test
│   ├── body
│   ├── head
│   ├── camera

Body

In gt.csv for body pose, for each line, the joint information is structured as follows:

x_2d_1, y_2d_1, x_2d_2, y_2d_2, ... x_2d_15, y_2d_15, x_3d_1, y_3d_1, z_3d_1, ..., x_3d_15, y_3d_15, z_3d_15

Head

In gt.csv for head pose, for each line, the rotation information is structured as follos:

pitch, yaw, roll

Camera

In gt.csv for camear pose, for each line, the rotation information is structured as follows (Please ignore the roll data.):

pitch, yaw, roll(dummy)

Citation

If you find our work useful in your research, please cite our paper.

@inproceedings{hwang20_uist,
    author = {Hwang, Dong-Hyun and Aso, Kohei and Yuan, Ye and Kitani, Kris and Koike, Hideki},
    title = {MonoEye: Multimodal Human Motion Capture System Using A Single Ultra-Wide Fisheye Camera},
    year = {2020},
    doi = {10.1145/3379337.3415856},
    booktitle = {Proceedings of the 33rd Annual ACM Symposium on User Interface Software and Technology},
    pages = {98–111},
    numpages = {14},
    location = {Virtual Event, USA},
    series = {UIST '20}
}