app.py

import gradio as gr
import torch
import torch.nn as nn
from torchvision import models, transforms
from PIL import Image
import json

# Define class names
class_names = ['acorn_squash', 'almond', 'almonds', 'anchovy_(fish)', 'apple', 'apricot', 'artichoke', 'arugula', 'asparagus', 'avocado', 'baguette', 'banana', 'barley', 'barley_(grain)', 'beef_(meat)', 'beet', 'black_beans', 'black_pepper_(spice)', 'blackberry', 'bok_choy', 'bread_(loaf)', 'breadcrumbs', 'broccoli', 'brussels_sprouts', 'butter_(dairy)', 'butternut_squash', 'cabbage', 'canola_oil', 'cantaloupe', 'carrot', 'cashews', 'cauliflower', 'celery', 'cheddar_cheese', 'cherry', 'chicken_(meat)', 'chickpeas', 'chive', 'chocolate_chips', 'clams', 'clams_(seafood)', 'cocoa_powder', 'coconut', 'cod_(fish)', 'condensed_milk', 'confectioners_sugar', 'corn', 'corn_syrup', 'cornflakes', 'cornmeal', 'cottage_cheese', 'crab_(seafood)', 'crackers', 'cranberry', 'cream_(dairy)', 'cream_cheese', 'cucumber', 'date_(fruit)', 'dragonfruit', 'duck_(meat)', 'egg', 'eggplant', 'evaporated_milk', 'feta', 'feta_cheese', 'fig', 'fish_sauce', 'garlic', 'goat_cheese', 'grape', 'ground_beef', 'ground_pork', 'ground_turkey', 'guava', 'honeydew', 'jackfruit', 'kale', 'ketchup', 'kidney_beans', 'kiwi_(fruit)', 'lamb_(meat)', 'leek', 'lemon_(fruit)', 'lentils', 'lettuce', 'lime_(fruit)', 'lobster_(seafood)', 'lychee', 'mango', 'mayonnaise', 'meatballs', 'milk_(dairy)', 'molasses', 'mozzarella_cheese', 'mulberry', 'mushroom', 'mussels_(seafood)', 'mustard_(condiment)', 'mustard_greens', 'navy_beans', 'nectarine', 'noodles_(cooked)', 'oats', 'oats_(grain)', 'octopus_(seafood)', 'okra', 'olive', 'olive_oil', 'onion', 'orange_(fruit)', 'oyster_sauce', 'papaya', 'parmesan_cheese', 'parsnip', 'passionfruit', 'pasta_(cooked)', 'peach', 'peanut', 'peanut_butter', 'pear', 'pecans', 'pepper', 'persimmon', 'pineapple', 'pinto_beans', 'pita_bread', 'plum', 'pomegranate', 'pork_(meat)', 'potato', 'powdered_milk', 'powdered_sugar', 'pumpkin_seeds', 'quinoa', 'quinoa_(grain)', 'radish', 'raspberry', 'rice_(brown,_grain)', 'rice_(white,_grain)', 'ricotta', 'ricotta_cheese', 'rolled_oats', 'salmon', 'salmon_(fish)', 'salt', 'sardine_(fish)', 'scallion', 'scallops_(seafood)', 'seitan', 'sesame_oil', 'sesame_seeds', 'shallot', 'shrimp_(seafood)', 'sour_cream', 'soy_sauce', 'spinach', 'split_peas', 'squid_(seafood)', 'starfruit', 'strawberry', 'sunflower_oil', 'sunflower_seeds', 'sweet_potato', 'swiss_chard', 'tangerine', 'tempeh', 'tofu', 'tomato', 'tortilla_(flatbread)', 'tortillas', 'tuna', 'tuna_(fish)', 'turkey_(meat)', 'turnip', 'vegetable_oil', 'walnut', 'walnuts', 'watermelon', 'wheat_flour', 'whipping_cream', 'yam', 'yogurt_(dairy)', 'zucchini']

# Load model
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(f"Using device: {device}")

model = models.mobilenet_v2(pretrained=False)
num_ftrs = model.classifier[1].in_features
model.classifier[1] = nn.Linear(num_ftrs, len(class_names))
model.load_state_dict(torch.load("ingredient_recognition_model.pth", map_location=device))
model = model.to(device)
model.eval()

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

def clean_ingredient_name(name):
    """Clean up ingredient name for display"""
    name = name.split('_(')[0]
    name = name.replace('_', ' ')
    return name.title()

def predict(image):
    """Predict ingredients from image"""
    if image is None:
        return {"error": "No image provided"}
    
    try:
        # Convert to PIL Image if needed
        if not isinstance(image, Image.Image):
            image = Image.fromarray(image)
        
        # Convert to RGB
        image = image.convert("RGB")
        
        # Transform image
        input_tensor = transform(image).unsqueeze(0).to(device)
        
        # Run inference
        with torch.no_grad():
            outputs = model(input_tensor)
            probs = torch.nn.functional.softmax(outputs[0], dim=0)
            
            # Get top 5 predictions
            k = min(5, len(class_names))
            top_probs, top_idxs = torch.topk(probs, k)
        
        # Build results dictionary
        results = {}
        for prob, idx in zip(top_probs, top_idxs):
            raw_name = class_names[idx]
            clean_name = clean_ingredient_name(raw_name)
            confidence = prob.item()
            results[clean_name] = float(confidence)
        
        return results
        
    except Exception as e:
        return {"error": str(e)}

# Create Gradio interface
iface = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="pil", label="Upload Food Image"),
    outputs=gr.Label(num_top_classes=5, label="Predictions"),
    title="Ingredient Classification Model",
    description="Upload an image of food to identify the ingredients. This model recognizes over 200 different foods and ingredients.",
    examples=[
        # Add example images if you have them
    ],
    theme=gr.themes.Soft(),
    allow_flagging="never"
)

if __name__ == "__main__":
    iface.launch()