linfeng302 commited on
Commit
ced41e6
·
verified ·
1 Parent(s): 90315b7

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +65 -5
README.md CHANGED
@@ -1,6 +1,51 @@
1
- # UniBallRacket Dataset
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2
 
3
- This directory contains the dataset for the UniBallRacket framework, covering three racket sports: **badminton**, **table tennis**, and **tennis**.
 
 
 
 
4
 
5
  ## Directory Layout
6
 
@@ -22,10 +67,10 @@ data/
22
  │ │ └── <match>_<rally>.mp4 # Raw video clips
23
  │ ├── all/
24
  │ │ └── <match>/
25
- │ │ ├── median.npz # Median background frame (for BallTrack)
26
- │ │ ├── frame/<rally>/ # Extracted JPG frames
27
  │ │ ├── csv/<rally>_ball.csv # Ball ground truth annotations
28
  │ │ └── racket/<rally>/*.json # Racket ground truth annotations
 
 
29
  │ └── info/
30
  │ ├── metainfo.json # Sport-specific metadata
31
  │ ├── train.json # [[match_id, rally_id], ...] for training
@@ -37,6 +82,8 @@ data/
37
  └── ball_racket_<sport>_h80_f20.pkl # Long-horizon: 80 history → 20 future
38
  ```
39
 
 
 
40
  ## Data Formats
41
 
42
  ### Ball Annotations (`csv/<rally>_ball.csv`)
@@ -117,7 +164,7 @@ JSON list of `[match_id, rally_id]` pairs:
117
 
118
  ## Generating Data from Scratch
119
 
120
- If you have the raw videos, use `DataPreprocess/` scripts to prepare all intermediate files:
121
 
122
  ```bash
123
  cd DataPreprocess
@@ -151,3 +198,16 @@ python merge_gt_with_predictions.py --data_root ../data --sport badminton
151
  # Build PKL dataset
152
  python build_dataset.py --data_root ../data --sport badminton --history 80 --future 20
153
  ```
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ language:
4
+ - en
5
+ pretty_name: RacketVision Dataset
6
+ size_categories:
7
+ - 10K<n<100K
8
+ task_categories:
9
+ - object-detection
10
+ - video-classification
11
+ tags:
12
+ - sports-analytics
13
+ - computer-vision
14
+ - object-tracking
15
+ - trajectory-prediction
16
+ - ball-tracking
17
+ - racket-pose-estimation
18
+ - badminton
19
+ - table-tennis
20
+ - tennis
21
+ - racket-sports
22
+ ---
23
+
24
+ # RacketVision Dataset
25
+
26
+ [![Arxiv](https://img.shields.io/badge/ArXiv-2511.17045-B31B1B.svg)](https://arxiv.org/abs/2511.17045)
27
+ [![AAAI](https://img.shields.io/badge/AAAI_2026-Oral-blue.svg)](https://aaai.org/)
28
+ [![GitHub](https://img.shields.io/badge/GitHub-Code-black?logo=github)](https://github.com/OrcustD/RacketVision/)
29
+ [![Hugging Face Dataset](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Dataset-blue)](https://huggingface.co/datasets/linfeng302/RacketVision)
30
+ [![Hugging Face Models](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Models-yellow)](https://huggingface.co/linfeng302/RacketVision-Models)
31
+ [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](https://github.com/OrcustD/RacketVision/blob/main/LICENSE)
32
+
33
+ **RacketVision** is a large-scale, multi-sport dataset and benchmark for advancing computer vision in sports analytics, covering **badminton**, **table tennis**, and **tennis**. It is the first dataset to provide large-scale, fine-grained annotations for racket pose alongside traditional ball positions, enabling research into complex human-object interactions. The benchmark tackles three interconnected tasks: fine-grained **ball tracking**, articulated **racket pose estimation**, and predictive ball **trajectory forecasting**.
34
+
35
+ - **Paper:** [arXiv:2511.17045](https://arxiv.org/abs/2511.17045)
36
+ - **Code:** [github.com/OrcustD/RacketVision](https://github.com/OrcustD/RacketVision/)
37
+
38
+ ![Teaser](https://raw.githubusercontent.com/OrcustD/RacketVision/main/assets/teaser.jpg)
39
+
40
+ ## Using this Hub repository
41
+
42
+ This dataset is distributed as **static files** (videos, CSV, JSON, PKL). Download it with the Hugging Face CLI, then follow the [project README](https://github.com/OrcustD/RacketVision/blob/main/README.md) for environment setup and training:
43
 
44
+ ```bash
45
+ hf download linfeng302/RacketVision --repo-type dataset --local-dir data
46
+ ```
47
+
48
+ The in-browser Dataset Viewer may not fully load all assets: COCO detection and pose JSON files use different annotation schemas, so they are not merged into a single `datasets`-style table. Use the files on disk as documented below.
49
 
50
  ## Directory Layout
51
 
 
67
  │ │ └── <match>_<rally>.mp4 # Raw video clips
68
  │ ├── all/
69
  │ │ └── <match>/
 
 
70
  │ │ ├── csv/<rally>_ball.csv # Ball ground truth annotations
71
  │ │ └── racket/<rally>/*.json # Racket ground truth annotations
72
+ │ ├── interp_ball/ # Interpolated ball trajectories (for rebuilding TrajPred data)
73
+ │ ├── merged_racket/ # Merged racket predictions (for rebuilding TrajPred data)
74
  │ └── info/
75
  │ ├── metainfo.json # Sport-specific metadata
76
  │ ├── train.json # [[match_id, rally_id], ...] for training
 
82
  └── ball_racket_<sport>_h80_f20.pkl # Long-horizon: 80 history → 20 future
83
  ```
84
 
85
+ **Local preprocessing (required for BallTrack):** after download, generate per-match `frame/<rally>/` (JPG frames) and `median.npz` from the videos using `DataPreprocess/extract_frames.py` and `DataPreprocess/create_median.py`. These are omitted from the Hub release to save space; see the [project README](https://github.com/OrcustD/RacketVision/blob/main/README.md).
86
+
87
  ## Data Formats
88
 
89
  ### Ball Annotations (`csv/<rally>_ball.csv`)
 
164
 
165
  ## Generating Data from Scratch
166
 
167
+ If you have the raw videos, use `DataPreprocess/` scripts in the [code repository](https://github.com/OrcustD/RacketVision/) to prepare all intermediate files:
168
 
169
  ```bash
170
  cd DataPreprocess
 
198
  # Build PKL dataset
199
  python build_dataset.py --data_root ../data --sport badminton --history 80 --future 20
200
  ```
201
+
202
+ ## Citation
203
+
204
+ If you find this work useful, please consider citing:
205
+
206
+ ```bibtex
207
+ @inproceedings{dong2026racket,
208
+ title={Racket Vision: A Multiple Racket Sports Benchmark for Unified Ball and Racket Analysis},
209
+ author={Dong, Linfeng and Yang, Yuchen and Wu, Hao and Wang, Wei and Hou, Yuenan and Zhong, Zhihang and Sun, Xiao},
210
+ booktitle={Proceedings of the AAAI Conference on Artificial Intelligence (AAAI)},
211
+ year={2026}
212
+ }
213
+ ```