Nikhil0702/Gender_Classification_Using_Hybrid_Approach

1import os2import pickle3import cv24import numpy as np5import streamlit as st6import matplotlib.pyplot as plt7from tensorflow.keras.models import load_model8 9# ---------------- SETTINGS ----------------10BASE_DIR = "Main_py"11 12# ---------------- FILE CHECK ----------------13def check_file(filename):14    path = os.path.join(BASE_DIR, filename)15    if not os.path.exists(path):16        st.error(f"❌ File '{filename}' not found in '{BASE_DIR}'")17        st.stop()18    return path19 20# ---------------- LOAD MODELS ----------------21@st.cache_resource22def load_all_models():23    extractor = load_model(check_file("feature_extractor.keras"))24    with open(check_file("svm_model.pkl"), "rb") as f:25        svm = pickle.load(f)26    with open(check_file("rf_model.pkl"), "rb") as f:27        rf = pickle.load(f)28    with open(check_file("xgb_model.pkl"), "rb") as f:29        xgb = pickle.load(f)30    return extractor, svm, rf, xgb31 32feature_extractor, svm_model, rf_model, xgb_model = load_all_models()33 34# Face detector for rejecting non-face inputs35face_cascade = cv2.CascadeClassifier(36    cv2.data.haarcascades + 'haarcascade_frontalface_default.xml'37)38 39# ---------------- PREDICTION FUNCTION ----------------40def predict_single_image(uploaded_file):41    # Reset file pointer and read bytes42    uploaded_file.seek(0)43    file_bytes = np.asarray(bytearray(uploaded_file.read()), dtype=np.uint8)44    img = cv2.imdecode(file_bytes, cv2.IMREAD_COLOR)45 46    if img is None:47        # Could not decode image48        blank = np.zeros((300, 300, 3), dtype=np.uint8)49        return None, blank, 0.050 51    img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)52    img_display = cv2.resize(img_rgb, (300, 300), interpolation=cv2.INTER_AREA)53 54    # --- Face detection ---55    gray = cv2.cvtColor(img_rgb, cv2.COLOR_RGB2GRAY)56    faces = face_cascade.detectMultiScale(gray, scaleFactor=1.1, minNeighbors=5)57 58    if len(faces) == 0:59        return None, img_display, 0.0  # Reject if no face detected60 61    # --- Preprocess for model ---62    img_model = cv2.resize(img_rgb, (128, 128))63    img_model = cv2.cvtColor(img_model, cv2.COLOR_RGB2GRAY) / 255.064    img_model = np.expand_dims(img_model, axis=-1)  # (128, 128, 1)65    img_model = np.expand_dims(img_model, axis=0)   # (1, 128, 128, 1)66 67    # --- Extract features ---68    feature_vector = feature_extractor.predict(img_model)69    feature_vector = feature_vector.reshape(1, -1)70 71    # --- Get probabilities from models ---72    def safe_proba(model):73        try:74            return model.predict_proba(feature_vector)[0]75        except:76            pred = model.predict(feature_vector)[0]77            return [1.0 - pred, pred] if pred in [0, 1] else [0.5, 0.5]78 79    prob_svm = safe_proba(svm_model)80    prob_rf = safe_proba(rf_model)81    prob_xgb = safe_proba(xgb_model)82 83    # Average probabilities84    avg_probs = np.mean([prob_svm, prob_rf, prob_xgb], axis=0)85    confidence = float(np.max(avg_probs))86    predicted_class = int(np.argmax(avg_probs))87    gender_final = "Male" if predicted_class == 1 else "Female"88 89    # --- Reject if too low confidence ---90    if confidence < 0.70:91        return None, img_display, confidence92 93    return gender_final, img_display, confidence94 95# ---------------- STREAMLIT UI ----------------96st.set_page_config(page_title="Gender Classification App", page_icon="👤")97 98st.title("👤 Gender Classification (Hybrid DL + ML)")99st.write("Only predicts for **clear human male/female faces** — others will be rejected but still shown.")100 101uploaded_file = st.file_uploader("📤 Upload an image", type=["jpg", "jpeg", "png"])102 103if uploaded_file is not None:104    if st.button("🔍 Predict Gender"):105        gender, img_display, conf = predict_single_image(uploaded_file)106 107        fig, ax = plt.subplots()108        ax.imshow(img_display)109        ax.axis("off")110 111        if gender is None:112            msg = "❌ This is not detected as a picture of a man or woman."113            ax.set_title(msg, fontsize=10, color="red")114            st.pyplot(fig)115            st.warning(msg)116        else:117            color = "blue" if gender == "Male" else "green"118            ax.set_title(f"{gender} ({conf*100:.1f}% confident)", fontsize=12, color=color)119            st.pyplot(fig)120            st.success(f"✅ Predicted Gender: {gender} — {conf*100:.1f}% confidence")121