Image Seed Data Set

The Image Seed Data Set contains 276 real X-ray images of wheat kernels from three species: Kama (72 samples), Canadian (96 samples), and Rosa (108 samples). The images were obtained using a short-focus X-ray apparatus (Elektronika 25), digitized, and standardized by centering each grain image in a black-padded canvas of uniform size.

The dataset reflects class imbalance common in real-world data. To support robust evaluation, it includes five cross-validation (CV) splits. Each split consists of 80% training and 20% testing data, with disjoint test sets across folds. This means that test images in one fold are never reused in another.

Cross Validation Breakdown

The following table shows the number of training and test samples per class across all five folds:

Class	Set	Cross Validation Subset
Class	Set	Fold 0	Fold 1	Fold 2	Fold 3	Fold 4
Kama	Train	57	56	59	55	61
Kama	Test	15	16	13	17	11
Canadian	Train	81	77	69	84	73
Canadian	Test	15	19	27	12	23
Rosa	Train	83	88	93	82	86
Rosa	Test	25	20	15	26	22

Download Datasets

The following resources will be made available after the official article publication:

Full Image Set: X-ray images of all 276 samples.
5-Fold Cross-Validation Split: Stratified training/testing sets for reproducible experimentation.

Citation

If you use the dataset, please cite the following work:

Piotr A. Kowalski, Ernest Jeczmionek, Malgorzata Charytanowicz, Szymon Lukasik, and Piotr Kulczycki, Seeds Image - Introduction and Baseline Experiments with the New Labeled Benchmark for Machine Learning Tasks, Mobile Networks and Applications (2025). https://doi.org/10.1007/s11036-025-02479-0. LINK to paper
Piotr A. Kowalski, et al., Classification of Wheat Species Using Convolutional Neural Networks: A Comparative Study. In: Perakovic, D., Knapcikova, L. (eds) Future Access Enablers for Ubiquitous and Intelligent Infrastructures. FABULOUS 2024. Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering, vol 596. Springer, Cham. https://doi.org/10.1007/978-3-031-72393-3_1 LINK to paper

BibTeX


@article{KOWALSKI_seeds_2025:MONA, 

  	author = {Kowalski, Piotr A. and Jeczmionek, Ernest and Charytanowicz, Malgorzata and {\L}ukasik, Szymon and Kulczycki, Piotr}, 

  	doi = {10.1007/s11036-025-02479-0}, 

  	isbn = {1572-8153}, 

  	journal = {Mobile Networks and Applications}, 

  	title = {Seeds Image --Introduction and Baseline Experiments with the New Labeled Benchmark for Machine Learning Tasks}, 

  	url = {https://doi.org/10.1007/s11036-025-02479-0}, 

  	year = {2025}, 

  	bdsk-url-1 = {https://doi.org/10.1007/s11036-025-02479-0} 

} 
    
    
@InProceedings{kowalski2025classification,

  author = {Kowalski, Piotr A. and Jeczmionek, Ernest and Charytanowicz, Malgorzata and {\L}ukasik, Szymon and Niewczas, Jerzy and Kulczycki, Piotr},

  editor = {Perakovic, Dragan and Knapcikova, Lucia},

  title = {Classification of Wheat Species Using Convolutional Neural Networks: A Comparative Study},

  booktitle = {Future Access Enablers for Ubiquitous and Intelligent Infrastructures},

  year = {2025},

  publisher = {Springer Nature Switzerland},

  address = {Cham},

  pages = {3--8},

  isbn = {978-3-031-72393-3}

}