Repository of Image Databases

Introduction

Welcome to the Repository of Image Databases, here you can find ready to use, with your classifier, image databases with augmented image features and software to create your own augmented image databases.

Key Features

1. Image databases, CIFAR, COIL100, ISIC20XX, MNIST, and OCID, with augmented image features.
2. Sample images and statistics to validate that training CNNs with image databases with augmented features (singular points) boosts machine learning classification.
3. Software that was applied to generate the databases with augmented features.

Recent Additions

• Two versions of the Oral Cancer Image Database (OCID) [17] embedded with vector field (VF) ̅∇ψ̂ sized 224×224 and 400×400.
• OCID database with embedded new VF ̅∇g₁ = ∇(k₁ϕ̂ + k₂ψ̂) strings of singularities.
• New section with information about distilled image features along with sample images.

Contributors

Please visit the department webpage for more information about Mathematics at East Texas A&M University.

Main Contributors

Collaborators

a. East Texas A&M University

b. L2TI, University Sorbonne Paris Nord, France

Downloads

Software

• ELPAC-S
  • Documentation for ELPAC-S
• SRWC (Sparce Representation Wavelet Classification) – Used in [8] for classification of augmented image databases.

Databases

Augmented Real Valued Shapes / Singular Points (based on [8])

• COIL100
  • COIL100 – ∇ϕ̂
  • COIL100 – ∇ψ̂
• digit-MNIST [12] Training Images
  • digit-MNIST- ∇ψ̂
• digit-MNIST Testing Images
  • digit-MNIST – ∇ψ̂
• ISIC 2018 [10, 11]
  • ISIC2018 – ∇ϕ̂
  • ISIC2018 – ∇ψ̂
• ISIC 2020 Training Images
  • ISIC2020 – ∇ϕ̂
  • ISIC2020 – ∇ϕ̂ (VF Only)
  • ISIC2020 – ∇ψ̂
  • ISIC2020 – ∇ψ̂ (VF Only)
• ISIC 2020 Testing Images
  • ISIC2020 – ∇ϕ̂
  • ISIC2020 – ∇ϕ̂ (VF Only)
  • ISIC2020 – ∇ψ̂
  • ISIC2020 – ∇ψ̂ (VF Only)

Augmented Complex Valued Shapes / Singular Points (based on [6]) *

• CIFAR10
  • CIFAR10 – ̅∇ψ̂
• COIL100
  • COIL100 – ̅∇û
  • COIL100 – noise – ̅∇û
  • COIL100 – ̅∇ϕ̂
  • COIL100 – noise – ̅∇ϕ̂
  • COIL100 – ̅∇ψ̂
• digit-MNIST Training Images
  • digit-MNIST – ̅∇ψ̂
• digit-MNIST Testing Images
  • digit-MNIST – ̅∇ψ̂
• fashion-MNIST Training & Testing Images
  • fashion-MNIST – ̅∇ϕ̂
  • fashion-MNIST – ̅∇ψ̂
• ISIC 2020 Training Images
  • ISIC2020 – ̅∇û
  • ISIC2020 – noise – ̅∇û
  • ISIC2020 – ̅∇ϕ̂
  • ISIC2020 – noise – ̅∇ϕ̂
  • ISIC2020 – ̅∇ψ̂
• OCID
  • OCID [17] – ̅∇ψ̂ – 224×224
  • OCID [17] – ̅∇ψ̂ – 400×400

SPR Contour Lines (reference TBD)

• OCID
  • OCID – ̅∇g₁ = ∇(k₁ϕ̂ + k₂ψ̂)

Original Databases

• CIFAR-10 Original Image Databases
• COIL-100 Original Image Database

• ISIC 2018 Original Databases (Training and Testing Images)
• ISIC 2018 Original Databases (Training and Testing Images) Resized to 500×500
• ISIC 2020 Original Databases (Training and Testing Images)
• ISIC 2020 Original Databases Preprocessed with Laplace Operator (Training and Testing Images):
  • 256×192
  • 590×332

• digit-MNIST [12]

• fashion-MNIST [13]
• fashion-MNIST Resized to 400×400

• OCID [17] Original Database
• OCID Original Database Resized to 224×224
• OCID Original Database Resized to 400×400

* In papers [6, 8, 9] we used v_û, v_ϕ̂, and v_ψ̂ to denote the complex valued VFs. Now we use ̅∇û, ̅∇ϕ̂, and ̅∇ψ̂, respectively. These new notations are also used within the ELPAC software.

Augmented Image Features

Below are some sample image from the repository along with results from the related research demonstrating the effectiveness of augmenting database images with more features. See papers [6, 8, 9] listed in the references and the table below.

Sample Images

You may click on any image for an enlarged version.

COIL-100

ISIC2020

Fashion-MNIST

Metrics

	ISIC 2020 [1] Accuracy (%)		COIL-100 [7] Accuracy (%)
Model	Original / ̅∇û Augmented Features [6]	Original / ̅∇ψ̂ Augmented Features [15]	Original / ̅∇û Augmented Features [6]	Original / ̅∇ψ̂ Augmented Features [16]
DenseNet 121 [2]	89.40 / 88.88		88.72 / 87.88
ResNet 50 [3]	90.20 / 91.19	88.30 / 93.00	89.51 / 90.16	90.13 / 92.33
Inception ResNet V2 [4]	89.60 / 92.50		88.91 / 91.45
EfficientNet B0 [5]	91.69 / 92.90		90.99 / 91.85
SeyNet [6]	90.10 / 93.25		89.41/ 91.27

Distilled Image Features

Click here to download a set of 20 images distilled from the training OCID database images.

Citing This Work

When using any results, software, or databases from this webpage, please cite the corresponding paper in which the results, software, or databases have been published (see reference section below), along with the following:

Sirakov NM, Bowden A. Image Databases with Features Augmented with Singular-Point Shapes to Enhance Machine Learning. Electronics. 2024; 13(16):3150. https://doi.org/10.3390/electronics13163150

N.M. Sirakov, A. Bowden, T. Wang, J. Gamez, Repository of Image Databases with Augmented Features, https://www.etamu.edu/projects/augmented-image-repository/?redirect=none

When using any databases linked here, cite the original databases [1, 5, 7, 10-14, 17] as well.

When using the ELPAC-S software, please also include this citation:

A. Bowden, N.M. Sirakov, Active Contour Directed by the Poisson Gradient Vector Field and Edge Tracking, Journal of Mathematical Imaging and Vision – Springer, 63:665–680, 2021, https://rdcu.be/cflaI, (2021)

When using the SRWC software, include citation [8].

Conclusion

Acknowledgments

We would like to thank the following individuals for their past and continuing support of this work:

• Former Department Chair Dr. Tingxiu Wang in the Department of Mathematics at East Texas A&M University.
• Mr. Jeremy Gamez, Chief Information Officer, and Mr. Jeff Faunce, Director, Center for IT Excellence at East Texas A&M University.
• The team from the Office of Marketing and Communications at East Texas A&M University.

Supporting Publications

See papers [6, 8, 9] by the contributors that validate that augmenting image features increases machine learning effectiveness.

References

• [1]. International Skin Imaging Collaboration. SIIM-ISIC 2020 Challenge Dataset. International Skin Imaging Collaboration https://doi.org/10.34970/2020-ds01 (2020). [Creative Commons Attribution-Non Commercial 4.0 International License. The dataset was generated by the International Skin Imaging Collaboration (ISIC) and images are from the following sources: Hospital Clínic de Barcelona, Medical University of Vienna, Memorial Sloan Kettering Cancer Center, Melanoma Institute Australia, The University of Queensland, and the University of Athens Medical School. You should have received a copy of the license along with this work. If not, see https://creativecommons.org/licenses/by-nc/4.0/legalcode.txt .]
• [2]. K. He, X. Zhang, S. Ren and J. Sun, “Deep residual learning for image recognition.” In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 2016 pp. 770-778. doi: 10.1109/CVPR.2016.90
• [3]. Huang, G., Liu, Z., Van Der Maaten, L., Weinberger, K.Q: Densely connected convolutional networks. IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (2017). pp. 22612269. doi: 10.1109/CVPR.2017.243
• [4]. C. Szegedy, S. Ioe, and V. Vanhoucke, Inception-v4, inception-resnet and the impact of residual connections on learning, Proc. of the 31st AAAI Conference on Artificial Intelligence (AAAI-17), pp.4278-4284, (2017)
• [5]. T. A. Putra, S.I Rufaida, J-S Leu, Enhanced Skin Condition Prediction Through Machine Learning Using Dynamic Training and Testing Augmentation, DOI:10.1109/ACCESS.2020.2976045, IEEE Access, (2020).
• [6]. Igbasanmi, O., Sirakov, N.M., Bowden, A., CNN for Efficient Objects Classification with Embedded Vectors Fields, in printing by the Springer book series, Studies in Computational Intelligence, Electronic ISSN 1860-9503, Print ISSN 1860-949X, Best Paper Award
• [7]. “Columbia Object Image Library (COIL-100),” S. A. Nene, S. K. Nayar and H. Murase, Technical Report CUCS-006-96, February 1996.
• [8]. N.M. Sirakov, A. Bowden, M. Chen, L.H. Ngo, M. Luong, Embedding vector field into image features to enhance classification, Journal of Computational and Applied Mathematics, Vol. 441, 2024, 115685, ISSN 0377-0427, https://doi.org/10.1016/j.cam.2023.115685, https://www.sciencedirect.com/science/article/pii/S0377042723006283
• [9]. Oluwaseyi Dotum Igbasanmy, Nikolay Metodiev Sirakov, On the Usefulness of the Vector Field Singular Points Shapes for Classification, International Journal of Applied and Computational Mathematics, Springer, Accepted for publication January 12, 2024, DOI: https://doi.org/10.21203/rs.3.rs-2862010/v1
• [10]. HAM10000 Dataset: (c) by ViDIR Group, Department of Dermatology, Medical University of Vienna; https://doi.org/10.1038/sdata.2018.161
• [11]. MSK Dataset: (c) Anonymous; https://arxiv.org/abs/1710.05006; https://arxiv.org/abs/1902.03368
• [12]. MNIST Dataset: https://github.com/cvdfoundation/mnist
• [13]. Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms. Han Xiao, Kashif Rasul, Roland Vollgraf. arXiv:1708.07747
• [14]. CIFAR-10 Dataset: https://www.cs.toronto.edu/~kriz/cifar.html
• [15]. Eluwumi F. Petrus-Nihi, Rumana Akther, Nikolay Metodiev Sirakov, Image Features and Image Dataset Augmentation for Skin Lesions Machine Learning Classification, MCO2025, Mentz, France, Special sessions Innovations for Health: From Acquisition to AI-Driven Analysis, June 4-6, 2025, Metz, France, In printing by “Lecture Notes in Networks and System-Springer-Artificial Intelligence”.
• [16]. Eluwumi Folake Petrus-Nihi, Nikolay Metodiev Sirakov, Embedding a New Conjugate Gradient Vector Field to Augment Image Features and Enhance Machine Learning Classification, INFUS2025, July 29-31, Istanbul, Turkey, 2025. In printing by “Lecture Notes in Networks and Systems – Springer – Intelligent and Fuzzy Systems”.
• [17]. Histopathologic Oral Cancer Detection using CNNs: https://www.kaggle.com/datasets/ashenafifasilkebede/dataset