VeritaDerm

VeritaDerm 🩺✨

📌 Overview

VeritaDerm is a high-performance computer vision model designed for the automated detection and classification of common dermatological conditions. Trained on a curated dataset of 5,000 images, VeritaDerm leverages the latest YOLO11 architecture to provide a balance between real-time inference speed and clinical accuracy.

This model is intended to assist in research and act as a preliminary screening tool for identifying dermatological patterns in digital imagery.

📊 Performance Metrics

The model achieved the following results on the validation set after rigorous training on an NVIDIA RTX A6000:

🧬 Supported Classes (8)

The model is trained to identify the following categories:

1. Acne
2. Chicken Skin (Keratosis Pilaris)
3. Eczema
4. Leprosy
5. Psoriasis
6. Ringworm
7. Warts
8. Healthy Skin (Background/Control)

🚀 How to Use

You can run VeritaDerm directly using the ultralytics library.

1. Install Requirements

pip install ultralytics

2. Run Inference

from ultralytics import YOLO

# Load the model from Hugging Face
model = YOLO("XythicK/veritaderm")

# Predict on an image
results = model.predict(source="path_to_skin_image.jpg", conf=0.25)

# View results
results[0].show()

🛠️ Training Details

• Hardware: NVIDIA RTX A6000

• Dataset Size: 5,000 high-resolution dermatological images.

• Optimizer: Auto (SGD/AdamW)

• Epochs: 42 (Optimized)

• Augmentations: Mosaic, Mixup, and HSV-adjustments used to enhance generalizability.

⚠️ Medical Disclaimer

VeritaDerm is provided for educational and research purposes only. It is NOT a substitute for professional medical advice, diagnosis, or treatment. Always seek the advice of a qualified dermatologist or healthcare provider with any questions you may have regarding a medical condition.

✉️ Contact & Citation

If you use this model in your research or project, please credit the author:

@misc{xythick2026veritaderm,
  author = {M Mashhudur Rahim},
  title = {VeritaDerm: A Diagnostic Framework for Multi-Class Skin Disease Detection},
  year = {2026},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/XythicK/veritaderm}}
}