app.py

import streamlit as st
from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM
from PIL import Image

# Load the image classification pipeline
@st.cache_resource
def load_image_classification_pipeline():
    """
    Load the image classification pipeline using a pretrained model.
    """
    return pipeline("image-classification", model="Shresthadev403/food-image-classification")

pipe_classification = load_image_classification_pipeline()

# Load Qwen tokenizer and model
@st.cache_resource
def load_qwen_model():
    """
    Load the Qwen/Qwen2.5-Coder-32B-Instruct model and tokenizer.
    """
    tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-Coder-32B-Instruct")
    model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-Coder-32B-Instruct", device_map="auto")
    return tokenizer, model

# Function to generate ingredients using Qwen
def get_ingredients_qwen(food_name, tokenizer, model):
    """
    Generate a list of ingredients for the given food item using the Qwen model.
    """
    prompt = f"List the main ingredients typically used to prepare {food_name}:"
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    outputs = model.generate(**inputs, max_new_tokens=50)
    return tokenizer.decode(outputs[0], skip_special_tokens=True).strip()

# Streamlit app
st.title("Food Image Recognition with Ingredients")

# # Add the provided image as a banner
# st.image("CTP_Project/IR_IMAGE", caption="Food Recognition Model", use_column_width=True)

# Sidebar for model information
st.sidebar.title("Model Information")
st.sidebar.write("**Image Classification Model**: Shresthadev403/food-image-classification")
st.sidebar.write("**LLM for Ingredients**: Qwen2.5-Coder-32B-Instruct")

# Upload image
uploaded_file = st.file_uploader("Choose a food image...", type=["jpg", "png", "jpeg"])

# Load the Qwen model and tokenizer
tokenizer, model = load_qwen_model()

if uploaded_file is not None:
    # Display the uploaded image
    image = Image.open(uploaded_file)
    st.image(image, caption="Uploaded Image", use_column_width=True)
    st.write("Classifying...")

    # Make predictions
    predictions = pipe_classification(image)

    # Display only the top prediction
    top_food = predictions[0]['label']
    st.header(f"Food: {top_food}")

    # Generate and display ingredients for the top prediction
    st.subheader("Ingredients")
    try:
        ingredients = get_ingredients_qwen(top_food, tokenizer, model)
        st.write(ingredients)
    except Exception as e:
        st.error(f"Error generating ingredients: {e}")