Visual Coin-Tracking Dataset

The CTR dataset contains 17 sequences of coin-like objects annotated by visible side identification and object segmentation introduced in [1].

Previews of the dataset sequences are available at Vimeo, previews of the CTR-Base baseline method here.

Download

The dataset can be downloaded here: ctr.tar.gz (488MB). All of the annotations are licensed under CC BY-NC 4.0 (full license text in the CTR archive).

The dataset is organized as follows:

images/%08d.jpg - video frames
segmentations/%08d.png - corresponding ground truth segmentation masks
init_frames/sequence.json - names of the initialization frames from both sides
side_annotation/sequence.json - visible side annotation
bboxes/sequence.json - bounding boxes in [tl_x, tl_y, br_x, br_y] format (minimum axis-aligned bounding boxes of the segmentation ground truth)

The raw results of our CTR-Base method can be downloaded here: results.tar.gz (16MB)

Publication

Please cite the following paper if you use the dataset.

[1] J. Šerých and J. Matas,
Visual Coin-Tracking: Tracking of Planar Double-Sided Objects,
German Conference on Pattern Recognition. 2019

[PDF] [bib]

@inproceedings{serych2019cointracking,
  title = {Visual Coin-Tracking: Tracking of Planar Double-Sided Objects},
  author = {\v{S}er\'{y}ch, Jon\'{a}\v{s} and Matas, Ji\v{r}\'{i}},
  booktitle={German Conference on Pattern Recognition},
  pages={317--330},
  year={2019},
  organization={Springer}
}