OSAI introduces OpenTTGames - an open dataset aimed at evaluation of different computer vision tasks in Table Tennis: ball detection, semantic segmentation of humans, table and scoreboard and fast in-game events spotting.
It includes full-HD videos of table tennis games recorded at 120 fps with an industrial camera. Every video is equipped with an annotation, containing the frame numbers and corresponding targets for this particular frame: manually labeled in-game events (ball bounces, net hits, or empty event targets) and/or ball coordinates and segmentation masks, which were labeled with deep learning-aided annotation models.
The dataset consists of 5 videos from 10 to 25 min intended for training and a set of 7 short videos for testing. Each video is accompained with events and ball coordinates markup files and a folder with segmentation masks.
Markup file with event information consists of frame numbers and corresponding event's name:
{ ... "123": "bounce", ... }
Markup file with ball information consists of frame numbers and x,y coordinates of the ball on this frame, where (-1,-1) denotes abscence of the ball:
{ ... "123": {"x":456, "y":789}, ... }
Folder with segmentation masks consists of .png files where name of each file denotes corresponding frame number
... 123.png ...
The data is event-centred. The videos contain 4271 manually annotated events of 3 classes (ball bounces, net hits, and empty events). In addition to the events annotation, sequences of 4 frames before each event and 12 frames after has ball coordinates annotations and segmentation masks, which were labeled with deep learning-aided annotation models. The semantic segmentation (humans, table and scoreboard classes with channel-wise encoding) masks are presented as separate .png files for each annotated frame. The link to the mask file is provided in the main markup file.
15/04/2020 Initial publication of the dataset
If you use this dataset in your research, please consider to cite this publication:
Roman Voeikov, Nikolay Falaleev, and Ruslan Baikulov TTNet: Real-time temporal and spatial video analysis of table tennis. The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2020, pp. 884-885 [Publication][Bibtex]