app.py

import gradio as gr
import numpy as np
import cv2
import os

# Load Haar cascade for face detection
face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + "haarcascade_frontalface_default.xml")

# Load precomputed model data
mean = np.load("mean.npy")
eigenfaces = np.load("eigenfaces.npy")
projected_faces = np.load("projected_faces.npy")
labels = np.load("labels.npy")

# Preprocess the image: detect face, crop, grayscale, resize, flatten
def preprocess(img):
    try:
        gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
        gray = cv2.equalizeHist(gray)  # Improve lighting
    except Exception as e:
        raise ValueError(f"Failed to convert to grayscale: {e}")

    # Detect face
    faces = face_cascade.detectMultiScale(gray, scaleFactor=1.1, minNeighbors=5)
    if len(faces) == 0:
        raise ValueError("❌ No face detected in the image.")

    # Crop the first face found
    x, y, w, h = faces[0]
    face_roi = gray[y:y+h, x:x+w]
    resized_face = cv2.resize(face_roi, (100, 100))

    return resized_face.flatten()

# Predict function
def predict(img):
    if img is None:
        return "❌ No image provided."

    try:
        img_vector = preprocess(img)
        img_centered = img_vector - mean
        img_projected = np.dot(eigenfaces.T, img_centered)

        # Calculate Euclidean distances
        distances = np.linalg.norm(projected_faces - img_projected, axis=1)
        min_index = np.argmin(distances)
        predicted_label = labels[min_index]
        confidence = distances[min_index]

        return f"✅ Match: Person {predicted_label} (distance = {confidence:.2f})"

    except Exception as e:
        return f"⚠️ Error: {str(e)}"

# Gradio interface
iface = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="numpy", label="Upload or Capture a Face"),
    outputs=gr.Textbox(label="Prediction Result"),
    title="Enhanced Eigenfaces Face Recognition",
    description="Improved with face detection and lighting normalization. Upload a clear frontal face image."
)

iface.launch()