EfficientNet-B3 Skin Disease Classifier
Overview
This repository provides an EfficientNet-B3 image classification model trained to recognize common skin diseases from clinical skin images.
The model is intended for research and educational purposes only and should not be used as a substitute for professional medical diagnosis.
Model Architecture
- Backbone: EfficientNet-B3
- Framework: TensorFlow / Keras
- Task: Multi-class Image Classification
Disease Classes
The model predicts one of the following classes:
- Eczema
- ACD
- Psoriasis
- Tinea
- Urticaria
- Folliculitis
- Insect Bite
- Acne
Dataset
The model was trained using a custom dataset constructed from the SCIN (Skin Condition Image Network) dataset and DermNet images. The collected images were manually curated and mapped into eight diagnostic categories.
Input
- RGB Image
- Image Size: 300 × 300 pixels
Output
The model returns the probability for each disease class.
Example:
| Disease | Probability |
|---|---|
| Eczema | 0.82 |
| Psoriasis | 0.10 |
| Tinea | 0.04 |
Files
| File | Description |
|---|---|
| efficientnet_b3.keras | Trained classification model |
| efficientnet_backbone.keras | EfficientNet backbone |
| label_mapping.json | Class index mapping |
| training_config.json | Training configuration |
Example
import tensorflow as tf
model = tf.keras.models.load_model("efficientnet_b3.keras")
Intended Use
This model is designed for:
- Academic research
- Computer Vision experiments
- Medical AI education
- Prototype applications
Limitations
- Not intended for clinical diagnosis.
- Performance depends on image quality.
- Predictions should always be interpreted by healthcare professionals.
Author
Chantaro Ntw
AI Engineer | Computer Vision | Medical AI
- Downloads last month
- -