README.md

---
license: cc-by-nc-4.0
task_categories:
- video-classification
language:
- en
tags:
- video
- skeleton
- real basketball gym
- amateur basketball player
size_categories:
- 10B<n<100B
---
## **Dataset Card for MultiSubjects**
### MultiSubjects Introduction
MultiSubjects is a multi-subject single-player basketball action dataset for amateur basketball action recognition. 

Please refer to [this paper](https://www.sciencedirect.com/science/article/abs/pii/S1077314224002741) for more details.

If you wish to access it, please contact us and specify your organization and purpose. 

Our email: qws[email protected]; [email protected].
### Languages
The class labels in the dataset are in English.

### Action labels
d(0): dribble

p(1): lay up

s(2): shoot


### Dataset Structure
The file: Subjects1000.
```json
{
  {-1(subjects ID lable)
   -d(action label)
    -1_d_1.mp4('id'_'action label'_'video number')
   -p
    ...
   -s
    ...}

    .
    .
    .

  {-1000(subjects ID lable)
   -d
    -1000_d_1.mp4
     ...
   -p
    ...
   -s
    ...}
}
```
The file: MultiSubjects_train_val_test, for video action recognition.
```json
{
  train
   -'id'_'actionlabel video'_'number'.mp4
    ...
  val
   ...
  test
   ...

  train.txt
   -'video_name.mp4' 'action_label'
  val.txt
   ...
  test.txt
   ...
}
```
The file: MultiSubjects_SA, for skeleton-based action recognition. 

We use [mmaction2](https://github.com/open-mmlab/mmaction2/blob/main/configs/skeleton/posec3d/custom_dataset_training.md) to extract 2D human keypoints and construct the dataset in the format of the NTU dataset, divided into training and validation sets.
```json
{
  train.pkl
  val.pkl
}
```
The file: key_points_csv, the coordinates of 33 keypoints detected frame-wise by BlazePose.
The structure of 'id'_'action label'_'video number'.csv
```json
{
  'Frame number'{1,2,3...}
  'Keypoint_0'{x, y, z}, 'keypoint_1'{x, y, z} ... 'keypoint_32'{x, y, z}
}
```
The file: MultiSubjects_person_GT.
```json
{
  'Video name'{'id'_'action label'_'video number'.mp4 ...}
  'Frame number'{1,2,3...}
  'Box'{<x1, y1, x2, y2> ...}
  'Action label'{0, 1, 2}
  'Joint'{0}
}
```
### Dataset Curators
Authors of [this paper](https://www.sciencedirect.com/science/article/abs/pii/S1077314224002741)
* Zhijie Han
* Wansong Qin
* Yalu Wang
* Qixiang Wang
* Yongbin Shi

### Citation Information
```json
@article{han2024multisubjects,
  title={MultiSubjects: A multi-subject video dataset for single-person basketball action recognition from basketball gym},
  author={Han, Zhijie and Qin, Wansong and Wang, Yalu and Wang, Qixiang and Shi, Yongbin},
  journal={Computer Vision and Image Understanding},
  pages={104193},
  year={2024},
  publisher={Elsevier}
}
```