app.py

import streamlit as st
from transformers import ViTForImageClassification, ViTImageProcessor
from utils import load_image_vit, predict_toxicity_vit, get_label
from PIL import Image
import torch
import io

def classify_image(uploaded_file):
    # Load pre-trained ViT model and processor
    model_name = "google/vit-base-patch16-224"
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    model = ViTForImageClassification.from_pretrained(model_name)
    processor = ViTImageProcessor.from_pretrained(model_name)

    # Modify for binary classification (toxic/non-toxic)
    try:
        model.classifier = torch.nn.Linear(model.classifier.in_features, 2)
        model.load_state_dict(torch.load("toxic_classifier.pth", map_location=device), strict=False)
    except FileNotFoundError:
        st.warning("Using pre-trained ImageNet weights. For toxic classification, upload toxic_classifier.pth.")
    model.to(device)

    # Process image and predict
    inputs = load_image_vit(uploaded_file, processor)
    prediction, probabilities = predict_toxicity_vit(model, inputs, device)
    label = get_label(prediction)

    return label, probabilities

def main():
    st.title("ToxiScan - Toxic Image Classifier")
    st.write("Upload an image to detect if it contains toxic content using a pre-trained Vision Transformer.")

    # File uploader
    uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "png", "jpeg"])

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

        # Process and predict
        with st.spinner("Analyzing..."):
            label, probabilities = classify_image(uploaded_file)

            # Display results
            st.subheader("Results")
            st.write(f"**Prediction:** {label}")
            st.write(f"**Confidence Scores:**")
            st.write(f"- Toxic: {probabilities[1]:.2%}")
            st.write(f"- Non-Toxic: {probabilities[0]:.2%}")

            # Bar chart for visualization
            st.bar_chart({"Toxic": probabilities[1], "Non-Toxic": probabilities[0]})

if __name__ == "__main__":
    main()