app.py

import os
import io
import base64
import torch
from flask import Flask, request, jsonify
from flask_cors import CORS
from PIL import Image
from torchvision import transforms
from huggingface_hub import hf_hub_download

# Import model architecture
from model_architecture import create_model

# Initialize Flask app
app = Flask(__name__)
CORS(app)  # Enable CORS for all routes

# Constants
REPO_ID = "karthikeya09/smart_image_recognation"
MODEL_FILE = "best_model.pth"
CLASSES = ['glass', 'metal', 'non-recyclable', 'organic', 'paper', 'plastic']
CLASS_INFO = {
    'glass': {'emoji': '💚', 'color': '#28a745', 'description': 'Glass bottles, jars'},
    'metal': {'emoji': '🔘', 'color': '#6c757d', 'description': 'Cans, foil, batteries'},
    'non-recyclable': {'emoji': '⚫', 'color': '#343a40', 'description': 'Mixed/contaminated waste'},
    'organic': {'emoji': '🟢', 'color': '#20c997', 'description': 'Food waste, plant matter'},
    'paper': {'emoji': '📄', 'color': '#8B4513', 'description': 'Newspapers, cardboard'},
    'plastic': {'emoji': '🔵', 'color': '#007bff', 'description': 'Bottles, bags, containers'}
}

# Load Model
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print(f"📥 Downloading model from {REPO_ID}...")
model_path = hf_hub_download(repo_id=REPO_ID, filename=MODEL_FILE)

print("🧠 Loading model architecture...")
model = create_model(architecture='mobilenet_v2', num_classes=len(CLASSES), pretrained=False)
checkpoint = torch.load(model_path, map_location=device)
model.load_state_dict(checkpoint['model_state_dict'])
model = model.to(device)
model.eval()
print(f"✅ Model loaded on {device}")

# Transforms
transform = transforms.Compose([
    transforms.Resize((224, 224)),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])

def predict(img):
    """Run prediction on an image"""
    if img.mode != 'RGB':
        img = img.convert('RGB')
    
    input_tensor = transform(img).unsqueeze(0).to(device)
    
    with torch.no_grad():
        outputs = model(input_tensor)
        probabilities = torch.nn.functional.softmax(outputs, dim=1)
        confidence, predicted = torch.max(probabilities, 1)
    
    predicted_class = CLASSES[predicted.item()]
    confidence_score = confidence.item() * 100
    
    all_probs = {CLASSES[i]: round(probabilities[0][i].item() * 100, 2) 
                 for i in range(len(CLASSES))}
    
    return {
        'class': predicted_class,
        'confidence': round(confidence_score, 2),
        'info': CLASS_INFO[predicted_class],
        'all_probabilities': all_probs
    }

@app.route('/')
def home():
    return jsonify({
        'status': 'running',
        'model': REPO_ID,
        'classes': CLASSES,
        'message': 'Smart Waste Classifier API'
    })

@app.route('/predict', methods=['POST'])
def predict_endpoint():
    """Handle prediction requests"""
    try:
        # Handle JSON with base64 image
        if request.is_json:
            data = request.get_json()
            image_data = data.get('image_base64') or data.get('image')
            
            if not image_data:
                return jsonify({'success': False, 'error': 'No image provided'}), 400
            
            # Remove data URL prefix if present
            if ',' in image_data:
                image_data = image_data.split(',')[1]
            
            image_bytes = base64.b64decode(image_data)
            img = Image.open(io.BytesIO(image_bytes))
        
        # Handle form data with file upload
        elif 'image' in request.files:
            file = request.files['image']
            img = Image.open(file.stream)
        
        else:
            return jsonify({'success': False, 'error': 'No image provided'}), 400
        
        result = predict(img)
        return jsonify({'success': True, 'prediction': result})
    
    except Exception as e:
        return jsonify({'success': False, 'error': str(e)}), 500

@app.route('/health')
def health():
    return jsonify({'status': 'healthy', 'model_loaded': model is not None})

if __name__ == '__main__':
    app.run(host='0.0.0.0', port=7860)