PenileScreen-ViT

PenileScreen-ViT

Built upon:
➤ The Development and Performance of a Machine‑Learning Based Mobile Platform for Visually Determining the Etiology of 5 Penile Diseases — Allan‑Blitz LT, Ambepitiya S, Tirupathi R, & Klausner JD. Digital Health, 2024.
(Implementation and adaptation by our team.)

A Vision Transformer-based model for multi-class classification of penile-region dermatological images, focusing on visual patterns commonly associated with sexually transmitted conditions. Developed for research, academic study, and digital health tool prototyping.

🧠 Model Overview

The PenileScreen-ViT model categorizes input images into the following three visual classes:

• Genital_warts
• HSV (Herpes Simplex Virus)
• Syphilis

It is fine-tuned from google/vit-base-patch16-224-in21k using the TensorFlow and vit-keras frameworks and trained on a curated collection of de-identified dermatological images for academic and analytical purposes.

📦 Model Metadata

🧠 Model Architecture

This project uses:

• ViT-B16 pre-trained on ImageNet21k
• Custom classification head: Flatten -> Dense(3, softmax)
• Fine-tuned on a specialized, de-identified dataset of penile-region dermatological images
• Trained with educational and research use cases in mind

🎯 Purpose and Use

This model is intended for:

• Academic and AI research in visual pattern recognition
• Development of experimental digital health tools
• Exploration of visual features associated with selected STD-related dermatological cases
• Educational visualization in the field of medical AI and image classification

❗ This model is not intended for clinical use, diagnostic support, or real-world patient decision-making.

👨‍💻 Authors

• Janitha Prathapa
• Yudara Kularathne
• Thanveer Ahamad

📬 License

This project is licensed under the CC BY-NC 4.0 License.
Commercial use is prohibited without explicit permission. See the LICENSE file for details.

📚 Citation

BibTeX:

@misc{penilescreenvit2024,
  title={PenileScreen-ViT: Vision Transformer Model for STD-related Visual Classification},
  author={Yudara Kularathne, Janitha Prathapa and Thanveer Ahamad},
  year={2024},
  howpublished={\url{https://huggingface.co/HehealthVision/PenileScreen-ViT}},
}

Original paper (APA):

Allan‑Blitz LT, Ambepitiya S, Tirupathi R, & Klausner JD. (2024). The Development and Performance of a Machine‑Learning Based Mobile Platform for Visually Determining the Etiology of 5 Penile Diseases. Digital Health. Retrieved from https://www.mcpdigitalhealth.org/article/S2949-7612(24)00035-X/fulltext