app.py

import streamlit as st
from PIL import Image
import numpy as np
import joblib
import os
from skimage.feature import hog

IMAGE_SIZE = (128, 128)
MODEL_DIR = 'model'

# Load models
try:
    models = {
        'SVM': joblib.load(os.path.join(MODEL_DIR, 'svm.pkl')),
        'Random Forest': joblib.load(os.path.join(MODEL_DIR, 'rf.pkl')),
        'XGBoost': joblib.load(os.path.join(MODEL_DIR, 'xgb.pkl')),
        'KNN': joblib.load(os.path.join(MODEL_DIR, 'knn.pkl')),
    }
except FileNotFoundError as e:
    st.error(f"Model file not found: {e}")
    st.stop()

def extract_hog_features(image):
    image = image.resize(IMAGE_SIZE).convert('L')
    image_np = np.array(image)

    features = hog(image_np,
                   orientations=9,
                   pixels_per_cell=(8, 8),
                   cells_per_block=(2, 2),
                   block_norm='L2-Hys')

    if len(features) > 7056:
        features = features[:7056]

    return features.reshape(1, -1)

# UI
st.title("🧠 Autism Detection from Face Image")

uploaded_file = st.file_uploader("📤 Upload a face image", type=['jpg', 'jpeg', 'png'])

selected_model_name = st.selectbox("🔍 Choose a model for prediction:", list(models.keys()))

if uploaded_file is not None and selected_model_name:
    image = Image.open(uploaded_file)
    st.image(image, caption='Uploaded Image', use_container_width=True)

    with st.spinner(f'Extracting features and predicting with {selected_model_name}...'):
        try:
            features = extract_hog_features(image)
            model = models[selected_model_name]

            if selected_model_name == 'SVM':
                # SVM doesn't support predict_proba
                score = model.decision_function(features)[0]
                prediction = 1 if score >= 0 else 0
                st.info(f"📈 SVM decision score: {score:.2f}")
            else:
                prob = model.predict_proba(features)[0][1]
                prediction = 1 if prob >= 0.5 else 0
                st.info(f"📊 Predicted probability: {prob:.2f}")

            if prediction == 1:
                st.success("🔴 The model predicts **Autistic**.")
            else:
                st.success("🟢 The model predicts **Non-Autistic**.")

        except Exception as e:
            st.error(f"Error during prediction: {e}")