0xnu/skincare-detection

Skincare Disease Classification Model

A PyTorch-based deep learning model for classifying skincare diseases and conditions from images using transfer learning with EfficientNet-B0, ResNet50, and Vision Transformer (ViT).

Code Example

#!/usr/bin/env python3

import torch
from transformers import ViTImageProcessor, ViTForImageClassification
from PIL import Image
import json
from pathlib import Path
from typing import List, Dict, Union
import argparse

class SkincareClassifier:
    def __init__(self, model_name: str = '0xnu/skincare-detection'):
        self.processor = ViTImageProcessor.from_pretrained(model_name)
        self.model = ViTForImageClassification.from_pretrained(model_name)
        self.model.eval()
        self.id2label = self.model.config.id2label
    
    def classify(self, image_path: Union[str, Path], min_conf: float = 0.01) -> Dict:
        try:
            image = Image.open(image_path).convert('RGB')
            inputs = self.processor(images=image, return_tensors="pt")
            
            with torch.no_grad():
                outputs = self.model(**inputs)
                probs = torch.softmax(outputs.logits, dim=-1)[0]
            
            pred_id = outputs.logits.argmax().item()
            scores = {self.id2label[i]: float(probs[i]) for i in range(len(probs)) if probs[i] >= min_conf}
            
            return {
                'image': Path(image_path).name,
                'prediction': self.id2label[pred_id],
                'confidence': float(probs[pred_id]),
                'all_scores': dict(sorted(scores.items(), key=lambda x: x[1], reverse=True))
            }
        except Exception as e:
            return {'image': str(image_path), 'error': str(e)}
    
    def classify_batch(self, paths: List[Union[str, Path]], **kwargs) -> List[Dict]:
        return [self.classify(path, **kwargs) for path in paths]
    
    def classify_dir(self, dir_path: Union[str, Path], **kwargs) -> List[Dict]:
        extensions = {'.jpg', '.jpeg', '.png', '.bmp', '.tiff', '.webp'}
        paths = [p for p in Path(dir_path).rglob('*') if p.suffix.lower() in extensions]
        return self.classify_batch(paths, **kwargs) if paths else []
    
    def print_results(self, results: Union[Dict, List[Dict]]):
        if isinstance(results, dict):
            results = [results]
        
        for r in results:
            if 'error' in r:
                print(f"❌ {r['image']}: {r['error']}")
                continue
            
            print(f"📸 {r['image']}")
            print(f"🎯 {r['prediction'].upper()}: {r['confidence']:.1%}")
            
            for cls, conf in r['all_scores'].items():
                bar = "█" * int(conf * 20)
                print(f"   {cls:>8}: {conf:.1%} {bar}")
            print("-" * 30)

def main():
    parser = argparse.ArgumentParser(description="Skincare Image Classification")
    parser.add_argument('input', help='Image file or directory')
    parser.add_argument('--model', default='0xnu/skincare-detection')
    parser.add_argument('--output', help='JSON output file')
    parser.add_argument('--min-conf', type=float, default=0.01)
    args = parser.parse_args()
    
    classifier = SkincareClassifier(args.model)
    input_path = Path(args.input)
    
    if input_path.is_file():
        results = classifier.classify(input_path, args.min_conf)
    elif input_path.is_dir():
        results = classifier.classify_dir(input_path, min_conf=args.min_conf)
    else:
        return print(f"❌ Invalid path: {input_path}")
    
    classifier.print_results(results)
    
    if args.output:
        with open(args.output, 'w') as f:
            json.dump(results, f, indent=2, default=str)

if __name__ == "__main__":
    if len(__import__('sys').argv) == 1:
        classifier = SkincareClassifier()
        if Path('joe.jpeg').exists():
            classifier.print_results(classifier.classify('joe.jpeg'))
        else:
            print("Usage: python skincare.py <image_path>")
    else:
        main()

Execute Code

python skincare.py joe.jpeg

Limitations and Considerations

• Model performance depends on training data quality and diversity
• May not generalise well to significantly different image distributions
• Should not be used as sole diagnostic tool for medical decisions
• Requires validation by qualified healthcare professionals for clinical use
• Performance may vary across different skin types and demographics

Ethical Considerations

• This model is for educational and research purposes
• Medical applications require proper validation and regulatory approval
• Consider bias in training data and ensure diverse representation
• Implement appropriate safeguards for sensitive medical applications

Copyright

(c) Copyright 2025 Finbarrs Oketunji. All Rights Reserved.