import os os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3' # Suppress TensorFlow warnings import gradio as gr import tensorflow as tf from tensorflow.keras.models import load_model import numpy as np from PIL import Image import requests import json # ================== MODEL LOADING ================== try: model = load_model('wound_classifier_model_googlenet.h5') print("✅ Model loaded successfully") except Exception as e: raise RuntimeError(f"❌ Model loading failed: {str(e)}") # ================== CLASS LABELS ================== CLASS_LABELS = [ "Abrasions", "Bruises", "Burns", "Cut", "Diabetic Wounds", "Gingivitis", "Surgical Wounds", "Venous Wounds", "athlete foot", "cellulitis", "chickenpox", "cutaneous larva migrans", "impetigo", "nail fungus", "ringworm", "shingles", "tooth discoloration", "ulcer" ] # Verify model compatibility assert len(CLASS_LABELS) == model.output_shape[-1], \ f"Class mismatch: Model expects {model.output_shape[-1]} classes, found {len(CLASS_LABELS)}" # ================== OPENROUTER CONFIG ================== OPENROUTER_API_KEY = "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" OPENROUTER_API_URL = "https://openrouter.ai/api/v1/chat/completions" MODEL_NAME = "mistralai/mistral-7b-instruct" # ================== IMAGE PROCESSING ================== def preprocess_image(image, target_size=(224, 224)): """Process and validate input images""" try: if not image: raise ValueError("🚨 No image provided") image = image.convert("RGB").resize(target_size) array = np.array(image) / 255.0 print(f"🖼️ Image processed: Shape {array.shape}") return array except Exception as e: raise RuntimeError(f"🖼️ Image processing failed: {str(e)}") # ================== MEDICAL GUIDELINES ================== def get_medical_guidelines(wound_type): """Fetch treatment guidelines from OpenRouter API""" headers = { "Authorization": f"Bearer {OPENROUTER_API_KEY}", "Content-Type": "application/json", "HTTP-Referer": "https://huggingface.co/spaces/MahatirTusher/Wound_Treatment" } prompt = f"""You are a medical expert providing guidance for a patient with {wound_type}. First, briefly introduce what {wound_type} are and their typical characteristics. Then, provide a comprehensive care guide including: 1. Immediate First Aid Steps 2. Home Care Instructions 3. Prevention Tips 4. Warning Signs (when to seek emergency care) Format your response in clear sections with proper spacing. Use simple, non-technical language that a general audience can understand.""" try: print(f"📡 Sending API request for {wound_type}...") response = requests.post( OPENROUTER_API_URL, headers=headers, json={ "model": MODEL_NAME, "messages": [{"role": "user", "content": prompt}], "temperature": 0.5 }, timeout=20 ) response.raise_for_status() result = response.json() print("🔧 Raw API response:", json.dumps(result, indent=2)) if not result.get("choices"): return "⚠️ API response format unexpected" return result["choices"][0]["message"]["content"] except Exception as e: return f"⚠️ Guidelines unavailable: {str(e)}" # ================== MAIN PREDICTION ================== def predict(image): """Complete prediction pipeline""" try: # Preprocess image processed_img = preprocess_image(image) input_tensor = np.expand_dims(processed_img, axis=0) # Make prediction predictions = model.predict(input_tensor)[0] sorted_indices = np.argsort(predictions)[::-1] # Descending order # Format results results = { CLASS_LABELS[i]: float(predictions[i]) for i in sorted_indices[:3] # Top 3 predictions } top_class = CLASS_LABELS[sorted_indices[0]] # Get guidelines guidelines = get_medical_guidelines(top_class) return results, guidelines except Exception as e: return {f"🚨 Error": str(e)}, "" # ================== CSS STYLES ================== custom_css = """ .gradio-container { font-family: 'Times New Roman', Times, serif !important; } .container { background: rgba(255, 255, 255, 0.9); border-radius: 20px; padding: 2rem; box-shadow: 0 8px 32px 0 rgba(31, 38, 135, 0.37); backdrop-filter: blur(4px); border: 1px solid rgba(255, 255, 255, 0.18); margin: 1rem 0; } .title { color: #2D3748; text-align: center; font-size: 2.5rem; margin-bottom: 1rem; } .subtitle { color: #4A5568; text-align: center; font-size: 1.2rem; margin-bottom: 2rem; } .warning { background-color: #FED7D7; border-left: 4px solid #F56565; padding: 1rem; margin: 1rem 0; border-radius: 8px; } .info-section { background-color: #E6F6FF; border-radius: 12px; padding: 1.5rem; margin: 1rem 0; } .guidelines-box { background: #F7FAFC; border-radius: 10px; padding: 1rem; margin-top: 1rem; } """ # ================== GRADIO INTERFACE ================== def create_interface(): with gr.Blocks(css=custom_css, title="WoundWise AI") as demo: # Logo and Introduction with gr.Column(elem_classes="container"): gr.Image("logo.png", show_label=False, container=False) gr.Markdown("# Welcome to WoundWise AI") gr.Markdown(""" In today's fast-paced world, early detection of medical conditions can be a game-changer. WoundWise AI is a cutting-edge AI-powered wound classification system developed as part of EarlyMed, an initiative by students of Vellore Institute of Technology. Our mission is simple yet impactful: "Early Detection, Smarter Decision." By leveraging deep learning, this system analyzes wound images to provide accurate classifications and essential treatment guidelines. Whether it's a minor cut or a serious skin condition, our AI assists in identifying potential risks, helping users make informed health decisions at an early stage. """) # Main Interface with gr.Column(elem_classes="container"): gr.Markdown("## 📸 Upload or Capture Wound Image") gr.Markdown("Use your device's camera or upload an existing image") file_input = gr.Image(type="pil", label="Upload Wound Image") submit_btn = gr.Button("Analyze Now", variant="primary") # Example Images gr.Examples( examples=["abrasion.jpg", "burn.png", "bruise.png", "chicken-pox.png", "cut.png"], inputs=file_input, label="Example Images" ) output_label = gr.Label(label="Analysis Results", num_top_classes=3) output_guidelines = gr.Textbox( label="Medical Guidelines", lines=12, elem_classes="guidelines-box" ) # Why Care About Wounds with gr.Column(elem_classes="info-section"): gr.Markdown(""" ## ⚠️ Why Wound Care Matters Ignoring wounds can lead to serious complications: - Risk of bacterial infection - Delayed healing and scarring - Chronic wound development - Systemic health issues - Increased medical costs Early intervention and proper care are crucial for optimal healing and preventing complications. """) # How Our Model Works with gr.Column(elem_classes="info-section"): gr.Markdown(""" ## 🔬 Our Technology WoundWise AI uses a state-of-the-art deep learning model based on the GoogLeNet architecture, trained on thousands of medical images. The system: 1. Analyzes wound images using advanced computer vision 2. Identifies 18 different types of wounds and skin conditions 3. Provides real-time classification with confidence scores 4. Generates medical guidelines using advanced AI Our model achieves high accuracy through extensive training and validation by medical professionals. """) # Disclaimer with gr.Column(elem_classes="warning"): gr.Markdown(""" ## ⚕️ Medical Disclaimer This AI system is designed to assist in preliminary wound assessment only. It is not a substitute for professional medical advice, diagnosis, or treatment. Always seek the advice of qualified healthcare providers with any questions regarding your medical condition. If you believe your wound is serious or life-threatening, please seek immediate medical attention. """) # Connect input to processing submit_btn.click( fn=predict, inputs=[file_input], outputs=[output_label, output_guidelines] ) return demo if __name__ == "__main__": iface = create_interface() iface.launch( server_name="0.0.0.0", server_port=7860, share=True )