app.py

import streamlit as st
from ultralytics import YOLO
from huggingface_hub import hf_hub_download
from PIL import Image
import io
import cv2
import numpy as np

# Load YOLOv8 model from Hugging Face
repo_id = "Norphel/nutri-ai-n2"  # Replace with your Hugging Face model repo
model_file = "best.pt"
model_path = hf_hub_download(repo_id=repo_id, filename=model_file)
model = YOLO(model_path)

# Food nutrition data
foods = {
    "kewadatsi": {
        "calories": 400,
        "protein": "10g",
        "fat": "0.4g",
        "carbohydrates": "92g",
        "sodium": "20mg"
    },
    "emadatsi": {
        "calories": 275,
        "protein": "10g",
        "fat": "0.4g",
        "carbohydrates": "92g",
        "sodium": "850mg"
    },
    "rice": {
        "calories": 130,
        "protein": "2.7g",
        "fat": "0.3g",
        "carbohydrates": "28g",
        "sodium": "1mg"
    }
}

# Function to calculate bounding box area
def calculate_bounding_box_area(results):
    areas = []
    class_names = []

    # Iterate through all bounding boxes
    for box in results[0].boxes:
        x1, y1, x2, y2 = map(int, box.xyxy[0].tolist())  # Bounding box coordinates
        area = (x2 - x1) * (y2 - y1)  # Calculate area
        areas.append(area)
        
        # Get the class ID for the current bounding box
        class_ids = box.cls.tolist()  # This returns a list of class IDs
        print(class_ids)
        # Convert class ID(s) to class name(s)
        for class_id in class_ids:
            class_name = model.names[int(class_id)]  # Convert class ID to class name
            class_names.append(class_name)  # Add class name to list

    return class_names, areas


# Function to process the image (resize & convert)
def process_image(uploaded_file):
    if uploaded_file is not None:
        image = Image.open(uploaded_file)
        image = image.resize((640, 640))  # Resize to 640x640

        return image
    return None


# Function to run YOLO inference
def run_yolo(image):
    results = model(image)  # Run YOLOv8 detection with PIL image directly

    detected_classes, bounding_areas = calculate_bounding_box_area(results)

    # Draw bounding boxes on the image and return the processed image
    result_img = results[0].plot()  # Draw bounding boxes on the image
    result_pil = Image.fromarray(cv2.cvtColor(result_img, cv2.COLOR_BGR2RGB))  # Convert back to PIL
    print(detected_classes, bounding_areas)
    return result_pil, detected_classes, bounding_areas

# Streamlit UI
st.title("Nutri-AI")

uploaded_file = st.file_uploader("Upload an image (PNG, JPG, JPEG)", type=["png", "jpg", "jpeg"])

if uploaded_file:
    resized_image = process_image(uploaded_file)

    if resized_image:
        st.image(resized_image, caption="Resized Image (640x640)", use_container_width=True)
        st.write("🔍 Running YOLOv8 detection...")

        detected_image, detected_classes, bounding_areas = run_yolo(resized_image)
        print(detected_classes, bounding_areas)

        # Display detected image
        st.image(detected_image, caption="Detected Objects", use_container_width=True)

        # Show detected classes, bounding box areas, and food nutrition info
        for i, class_name in enumerate(detected_classes):
            st.subheader(f"🍲 Detected: {class_name.capitalize()}")
            st.write(f"🟡 **Bounding Box Area:** {bounding_areas[i]:,.2f} pixels²")

            if class_name in foods:
                st.write("📊 **Nutritional Information:**")
                for key, value in foods[class_name].items():
                    st.write(f"🔹 {key.capitalize()}: {value}")