Welcome to the GraniTR dataset, a comprehensive collection curated for the classification of granite slabs. This dataset was collected at the Granitaş Granite Factory in Bilecik, Turkey, with the aim of supporting the development of deep learning models for granite tile classification. It contains 934 images across six distinct granite types:
- Aksaray Yaylak
- Ankara Fume
- Balaban Green
- Crema Imperial
- Giresun Mink
- Hisar Gray
These images were captured under varying conditions (indoor/outdoor) and using different devices (Canon camera, Samsung Galaxy A50, and Huawei Mate 20 Pro) to ensure diversity in resolution, aspect ratio, and color. This variability is essential for training AI models that can effectively distinguish between granite types based on color, texture, and mineral composition.
The GraniTR dataset is intended to help both industry professionals and end-users identify granite slabs more accurately, enhancing the AI model's generalization ability for real-world applications.
- Total Images: 934
- Classes: 6 (Aksaray Yaylak, Ankara Fume, Balaban Green, Crema Imperial, Giresun Mink, Hisar Gray)
| Class Name | Total # Images | Sample Images |
|---|---|---|
| Aksaray Yaylak | 103 |  |
| Ankara Fume | 194 |  |
| Balaban Green | 114 |  |
| Crema Imperial | 188 |  |
| Giresun Mink | 264 |  |
| Hisar Gray | 71 |  |
To request access to the GraniTR dataset, please refer to the following research papers for more details:
- Impact of Image Augmentation on Deep Learning-Based Classification of Granite Tiles
IEEE Paper G. E. Bartos, S. Ünaldı and N. Yalçin, "Impact of Image Augmentation on Deep Learning-Based Classification of Granite Tiles," 2024 9th International Conference on Computer Science and Engineering (UBMK), Antalya, Turkiye, 2024, pp. 796-799, doi: 10.1109/UBMK63289.2024.10773433. - Explaining Deep Learning Decisions for Granite Tile Classification Using Feature Visualization
G. E. Bartos, S. Ünaldı, and N. Yalçın, "Explaining deep learning decisions for granite tile classification using feature visualization," 2025 IEEE 18th International Symposium on Computational Intelligence and Informatics (CINTI), Budapest, Hungary, Nov. 2025
Once you have reviewed the papers, you can request access to the dataset by contacting:
- Assistant Prof. Dr. Sibel Ünaldı
Email: sibel.unaldi@bilecik.edu.tr
Please refer to the dataset’s license information for terms of use and restrictions.