Turkish Lira Coin Classification Dataset

Turkish Coin Classification Dataset (TCDL)

This dataset was constructed to support object classification and instance segmentation research on Turkish coins. The raw collection consists of 684 original images captured using 3 different devices, featuring single, multiple, and overlapping coins placed on colorful, patterned, and irregular backgrounds. All coin instances were annotated in COCO format, yielding 3,550 annotations across 8 categories: 1 Kuruş, 5 Kuruş, 10 Kuruş, 25 Kuruş, 50 Kuruş, 1 Lira, 5 Lira, and Yabancı (foreign coin). In addition, 1,567 negative (background-only) samples ranging from 200 to 500 pixels were included to reduce false positive detections during inference.

Dataset Summary

Split
Images
Annotations

Original (coins)
684
3,550

Original (background)
1,567
0

Total
2,251
3,550

Class Distribution

Class
Count

10 Kuruş
511

1 Kuruş
363

1 Lira
509

25 Kuruş
474

50 Kuruş
438

5 Kuruş
416

5 Lira
456

Yabancı
383

The GitHub Link for the Graphical User Interface (GUI) used in the related to the article and this dataset: https://github.com/bbergsarr/Turkish_Lira_Coin_Counter_GUI

If you use this dataset in your research, please cite.

For more information, please contact:

www.linkedin.com/in/bertugcag

www.linkedin.com/in/buğra-berk-sarıoğlu

www.linkedin.com/in/gkgultekin