Update README with comprehensive usage instructions and Flask API examples
Browse files
README.md
CHANGED
@@ -7,48 +7,340 @@ tags:
|
|
7 |
- medical
|
8 |
- dermatology
|
9 |
- image-classification
|
|
|
|
|
|
|
10 |
library_name: keras
|
|
|
11 |
---
|
12 |
|
13 |
-
# DermaAI
|
14 |
|
15 |
-
|
16 |
|
17 |
-
|
18 |
|
19 |
-
|
20 |
|
21 |
-
|
|
|
|
|
|
|
|
|
22 |
|
23 |
-
|
24 |
-
|
|
|
|
|
25 |
- **Domain**: Medical/Dermatology
|
26 |
- **Framework**: TensorFlow/Keras
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
27 |
|
28 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
29 |
|
30 |
```python
|
|
|
|
|
31 |
import tensorflow as tf
|
|
|
|
|
|
|
|
|
|
|
32 |
from huggingface_hub import hf_hub_download
|
33 |
|
34 |
-
|
35 |
-
|
36 |
|
37 |
-
#
|
|
|
|
|
38 |
model = tf.keras.models.load_model(model_path)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
39 |
|
40 |
-
|
41 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
42 |
```
|
43 |
|
44 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
45 |
|
46 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
47 |
|
48 |
-
##
|
49 |
|
50 |
-
|
51 |
|
52 |
-
##
|
53 |
|
54 |
-
|
|
|
|
|
|
7 |
- medical
|
8 |
- dermatology
|
9 |
- image-classification
|
10 |
+
- skin-disease
|
11 |
+
- efficientnet
|
12 |
+
- healthcare
|
13 |
library_name: keras
|
14 |
+
pipeline_tag: image-classification
|
15 |
---
|
16 |
|
17 |
+
# DermaAI - Skin Disease Classification Model
|
18 |
|
19 |
+
A deep learning model for classifying skin diseases using computer vision. This model can identify 5 different skin conditions with confidence scores and medical recommendations.
|
20 |
|
21 |
+
## 🏥 Supported Skin Conditions
|
22 |
|
23 |
+
The model can classify the following skin diseases:
|
24 |
|
25 |
+
1. **Atopic Dermatitis** - A chronic inflammatory skin condition
|
26 |
+
2. **Eczema** - Inflammatory skin condition causing red, itchy patches
|
27 |
+
3. **Psoriasis** - Autoimmune condition causing scaly skin patches
|
28 |
+
4. **Seborrheic Keratoses** - Common benign skin growths
|
29 |
+
5. **Tinea Ringworm Candidiasis** - Fungal skin infections
|
30 |
|
31 |
+
## 🔧 Model Details
|
32 |
+
|
33 |
+
- **Model Type**: Keras/TensorFlow model based on EfficientNetV2
|
34 |
+
- **Task**: Image Classification (Multi-class)
|
35 |
- **Domain**: Medical/Dermatology
|
36 |
- **Framework**: TensorFlow/Keras
|
37 |
+
- **Input Size**: 224x224x3 (RGB images)
|
38 |
+
- **Output**: 5-class probability distribution
|
39 |
+
- **Preprocessing**: EfficientNetV2 preprocessing
|
40 |
+
|
41 |
+
## 🚀 Quick Start
|
42 |
+
|
43 |
+
### Basic Usage
|
44 |
+
|
45 |
+
```python
|
46 |
+
import tensorflow as tf
|
47 |
+
from huggingface_hub import hf_hub_download
|
48 |
+
import numpy as np
|
49 |
+
from PIL import Image
|
50 |
+
from tensorflow.keras.applications.efficientnet_v2 import preprocess_input
|
51 |
+
|
52 |
+
# Download and load the model
|
53 |
+
model_path = hf_hub_download(repo_id="Siraja704/DermaAI", filename="DermaAI.keras")
|
54 |
+
model = tf.keras.models.load_model(model_path)
|
55 |
+
|
56 |
+
# Class names
|
57 |
+
class_names = [
|
58 |
+
'Atopic Dermatitis',
|
59 |
+
'Eczema',
|
60 |
+
'Psoriasis',
|
61 |
+
'Seborrheic Keratoses',
|
62 |
+
'Tinea Ringworm Candidiasis'
|
63 |
+
]
|
64 |
+
|
65 |
+
# Prediction function
|
66 |
+
def predict_skin_condition(image_path):
|
67 |
+
# Load and preprocess image
|
68 |
+
image = Image.open(image_path).convert('RGB')
|
69 |
+
image = image.resize((224, 224))
|
70 |
+
image_array = np.array(image)
|
71 |
+
image_array = preprocess_input(image_array)
|
72 |
+
image_array = np.expand_dims(image_array, axis=0)
|
73 |
+
|
74 |
+
# Make prediction
|
75 |
+
predictions = model.predict(image_array)
|
76 |
+
predicted_class_index = np.argmax(predictions[0])
|
77 |
+
predicted_class = class_names[predicted_class_index]
|
78 |
+
confidence = predictions[0][predicted_class_index] * 100
|
79 |
+
|
80 |
+
return predicted_class, confidence
|
81 |
+
|
82 |
+
# Example usage
|
83 |
+
prediction, confidence = predict_skin_condition("path/to/your/image.jpg")
|
84 |
+
print(f"Prediction: {prediction} ({confidence:.2f}% confidence)")
|
85 |
+
```
|
86 |
+
|
87 |
+
## 🌐 Flask API Usage
|
88 |
+
|
89 |
+
Create a complete web API for skin disease classification:
|
90 |
|
91 |
+
### 1. Install Dependencies
|
92 |
+
|
93 |
+
```bash
|
94 |
+
pip install flask numpy tensorflow pillow flask-cors huggingface-hub
|
95 |
+
```
|
96 |
+
|
97 |
+
### 2. Create Flask Application (`app.py`)
|
98 |
|
99 |
```python
|
100 |
+
from flask import Flask, request, jsonify
|
101 |
+
import numpy as np
|
102 |
import tensorflow as tf
|
103 |
+
import base64
|
104 |
+
import io
|
105 |
+
from PIL import Image
|
106 |
+
from flask_cors import CORS
|
107 |
+
from tensorflow.keras.applications.efficientnet_v2 import preprocess_input
|
108 |
from huggingface_hub import hf_hub_download
|
109 |
|
110 |
+
app = Flask(__name__)
|
111 |
+
CORS(app)
|
112 |
|
113 |
+
# Download and load the model from Hugging Face
|
114 |
+
print("Downloading model from Hugging Face...")
|
115 |
+
model_path = hf_hub_download(repo_id="Siraja704/DermaAI", filename="DermaAI.keras")
|
116 |
model = tf.keras.models.load_model(model_path)
|
117 |
+
print("✅ Model loaded successfully!")
|
118 |
+
|
119 |
+
# Class names
|
120 |
+
class_names = [
|
121 |
+
'Atopic Dermatitis',
|
122 |
+
'Eczema',
|
123 |
+
'Psoriasis',
|
124 |
+
'Seborrheic Keratoses',
|
125 |
+
'Tinea Ringworm Candidiasis'
|
126 |
+
]
|
127 |
+
|
128 |
+
@app.route('/predict', methods=['POST'])
|
129 |
+
def predict():
|
130 |
+
try:
|
131 |
+
data = request.json
|
132 |
+
if not data or 'image' not in data:
|
133 |
+
return jsonify({'error': 'No image data provided'}), 400
|
134 |
+
|
135 |
+
# Process base64 image
|
136 |
+
image_data = data['image']
|
137 |
+
if 'base64,' in image_data:
|
138 |
+
image_data = image_data.split('base64,')[1]
|
139 |
+
|
140 |
+
# Decode and preprocess image
|
141 |
+
decoded_image = base64.b64decode(image_data)
|
142 |
+
image = Image.open(io.BytesIO(decoded_image)).convert('RGB')
|
143 |
+
image = image.resize((224, 224))
|
144 |
+
image_array = np.array(image)
|
145 |
+
image_array = preprocess_input(image_array)
|
146 |
+
image_array = np.expand_dims(image_array, axis=0)
|
147 |
+
|
148 |
+
# Make prediction
|
149 |
+
predictions = model.predict(image_array)
|
150 |
+
predicted_class_index = int(np.argmax(predictions[0]))
|
151 |
+
predicted_class = class_names[predicted_class_index]
|
152 |
+
confidence = float(predictions[0][predicted_class_index] * 100)
|
153 |
+
|
154 |
+
# Get top alternatives
|
155 |
+
top_indices = np.argsort(predictions[0])[-3:][::-1]
|
156 |
+
top_predictions = [
|
157 |
+
{
|
158 |
+
'class': class_names[i],
|
159 |
+
'confidence': float(predictions[0][i] * 100)
|
160 |
+
}
|
161 |
+
for i in top_indices if i != predicted_class_index
|
162 |
+
]
|
163 |
+
|
164 |
+
# Generate medical recommendation
|
165 |
+
if confidence < 10:
|
166 |
+
recommendation = "Very low confidence. Please retake image with better lighting and focus."
|
167 |
+
elif confidence < 30:
|
168 |
+
recommendation = "Low confidence. Preliminary result only. Consult a dermatologist."
|
169 |
+
elif confidence < 60:
|
170 |
+
recommendation = "Moderate confidence. Consider alternatives and consult healthcare professional."
|
171 |
+
else:
|
172 |
+
recommendation = "High confidence prediction. Always consult healthcare professional for confirmation."
|
173 |
|
174 |
+
return jsonify({
|
175 |
+
'prediction': predicted_class,
|
176 |
+
'confidence': round(confidence, 2),
|
177 |
+
'all_confidences': {
|
178 |
+
class_names[i]: float(pred * 100) for i, pred in enumerate(predictions[0])
|
179 |
+
},
|
180 |
+
'top_alternatives': top_predictions,
|
181 |
+
'recommendation': recommendation
|
182 |
+
})
|
183 |
+
|
184 |
+
except Exception as e:
|
185 |
+
return jsonify({'error': str(e)}), 500
|
186 |
+
|
187 |
+
@app.route('/health', methods=['GET'])
|
188 |
+
def health():
|
189 |
+
return jsonify({'status': 'healthy', 'model_loaded': True})
|
190 |
+
|
191 |
+
if __name__ == '__main__':
|
192 |
+
app.run(host='0.0.0.0', port=5001, debug=True)
|
193 |
```
|
194 |
|
195 |
+
### 3. Run the API
|
196 |
+
|
197 |
+
```bash
|
198 |
+
python app.py
|
199 |
+
```
|
200 |
+
|
201 |
+
The API will be available at `http://localhost:5001`
|
202 |
+
|
203 |
+
### 4. API Usage Examples
|
204 |
|
205 |
+
**Python Client:**
|
206 |
+
```python
|
207 |
+
import requests
|
208 |
+
import base64
|
209 |
+
|
210 |
+
def predict_image(image_path, api_url="http://localhost:5001/predict"):
|
211 |
+
with open(image_path, "rb") as image_file:
|
212 |
+
encoded_string = base64.b64encode(image_file.read()).decode('utf-8')
|
213 |
+
|
214 |
+
data = {"image": f"data:image/jpeg;base64,{encoded_string}"}
|
215 |
+
response = requests.post(api_url, json=data)
|
216 |
+
return response.json()
|
217 |
+
|
218 |
+
# Usage
|
219 |
+
result = predict_image("skin_image.jpg")
|
220 |
+
print(f"Prediction: {result['prediction']} ({result['confidence']}%)")
|
221 |
+
```
|
222 |
+
|
223 |
+
**JavaScript Client:**
|
224 |
+
```javascript
|
225 |
+
async function predictSkinCondition(imageFile) {
|
226 |
+
const base64 = await new Promise((resolve) => {
|
227 |
+
const reader = new FileReader();
|
228 |
+
reader.onload = () => resolve(reader.result);
|
229 |
+
reader.readAsDataURL(imageFile);
|
230 |
+
});
|
231 |
+
|
232 |
+
const response = await fetch('http://localhost:5001/predict', {
|
233 |
+
method: 'POST',
|
234 |
+
headers: {'Content-Type': 'application/json'},
|
235 |
+
body: JSON.stringify({image: base64})
|
236 |
+
});
|
237 |
+
|
238 |
+
return await response.json();
|
239 |
+
}
|
240 |
+
```
|
241 |
+
|
242 |
+
**cURL:**
|
243 |
+
```bash
|
244 |
+
curl -X POST http://localhost:5001/predict \
|
245 |
+
-H "Content-Type: application/json" \
|
246 |
+
-d '{"image": "data:image/jpeg;base64,YOUR_BASE64_IMAGE_HERE"}'
|
247 |
+
```
|
248 |
+
|
249 |
+
## 📋 API Response Format
|
250 |
+
|
251 |
+
```json
|
252 |
+
{
|
253 |
+
"prediction": "Eczema",
|
254 |
+
"confidence": 85.23,
|
255 |
+
"all_confidences": {
|
256 |
+
"Atopic Dermatitis": 12.45,
|
257 |
+
"Eczema": 85.23,
|
258 |
+
"Psoriasis": 1.32,
|
259 |
+
"Seborrheic Keratoses": 0.67,
|
260 |
+
"Tinea Ringworm Candidiasis": 0.33
|
261 |
+
},
|
262 |
+
"top_alternatives": [
|
263 |
+
{
|
264 |
+
"class": "Atopic Dermatitis",
|
265 |
+
"confidence": 12.45
|
266 |
+
}
|
267 |
+
],
|
268 |
+
"recommendation": "High confidence prediction. Always consult healthcare professional for confirmation."
|
269 |
+
}
|
270 |
+
```
|
271 |
+
|
272 |
+
## 🖼️ Image Requirements
|
273 |
+
|
274 |
+
- **Formats**: JPG, PNG, WebP, and other common formats
|
275 |
+
- **Size**: Automatically resized to 224x224 pixels
|
276 |
+
- **Quality**: High-resolution images with good lighting work best
|
277 |
+
- **Focus**: Ensure affected skin area is clearly visible
|
278 |
+
|
279 |
+
## 🐳 Docker Deployment
|
280 |
+
|
281 |
+
**Dockerfile:**
|
282 |
+
```dockerfile
|
283 |
+
FROM python:3.9-slim
|
284 |
+
|
285 |
+
WORKDIR /app
|
286 |
+
COPY requirements.txt .
|
287 |
+
RUN pip install -r requirements.txt
|
288 |
+
COPY app.py .
|
289 |
+
EXPOSE 5001
|
290 |
+
CMD ["python", "app.py"]
|
291 |
+
```
|
292 |
+
|
293 |
+
**Requirements.txt:**
|
294 |
+
```txt
|
295 |
+
flask>=2.0.0
|
296 |
+
numpy>=1.21.0
|
297 |
+
tensorflow>=2.13.0
|
298 |
+
pillow>=9.0.0
|
299 |
+
flask-cors>=3.0.0
|
300 |
+
huggingface-hub>=0.20.0
|
301 |
+
```
|
302 |
+
|
303 |
+
**Build and Run:**
|
304 |
+
```bash
|
305 |
+
docker build -t dermaai-api .
|
306 |
+
docker run -p 5001:5001 dermaai-api
|
307 |
+
```
|
308 |
+
|
309 |
+
## ⚕️ Important Medical Disclaimer
|
310 |
+
|
311 |
+
**This model is for educational and research purposes only. It should NOT be used as a substitute for professional medical diagnosis or treatment. Always consult qualified healthcare professionals for proper medical evaluation and treatment of skin conditions.**
|
312 |
+
|
313 |
+
## 📊 Performance Notes
|
314 |
+
|
315 |
+
- **Input**: 224x224 RGB images
|
316 |
+
- **Preprocessing**: EfficientNetV2 normalization
|
317 |
+
- **Architecture**: Based on EfficientNetV2
|
318 |
+
- **Classes**: 5 skin disease categories
|
319 |
+
- **Confidence Levels**:
|
320 |
+
- Low: < 30% (requires professional consultation)
|
321 |
+
- Moderate: 30-60% (consider alternatives)
|
322 |
+
- High: > 60% (still requires medical confirmation)
|
323 |
+
|
324 |
+
## 🤝 Citation
|
325 |
+
|
326 |
+
If you use this model in your research or applications, please cite appropriately:
|
327 |
+
|
328 |
+
```bibtex
|
329 |
+
@misc{dermaai2024,
|
330 |
+
title={DermaAI: Deep Learning Model for Skin Disease Classification},
|
331 |
+
author={Siraja704},
|
332 |
+
year={2024},
|
333 |
+
publisher={Hugging Face},
|
334 |
+
url={https://huggingface.co/Siraja704/DermaAI}
|
335 |
+
}
|
336 |
+
```
|
337 |
|
338 |
+
## 📝 License
|
339 |
|
340 |
+
Licensed under the Apache 2.0 License. See the LICENSE file for details.
|
341 |
|
342 |
+
## 🔗 Links
|
343 |
|
344 |
+
- **Model Repository**: [Siraja704/DermaAI](https://huggingface.co/Siraja704/DermaAI)
|
345 |
+
- **Framework**: [TensorFlow](https://tensorflow.org)
|
346 |
+
- **Base Architecture**: [EfficientNetV2](https://arxiv.org/abs/2104.00298)
|