import gradio as gr
import numpy as np
import cv2
import insightface
from insightface.app import FaceAnalysis

# Global variable to store the InsightFace app instance
app = FaceAnalysis(providers=['CPUExecutionProvider'])
app.prepare(ctx_id=0, det_size=(640, 640))

def liveness_check_passive(image):
    """
    Checks if the face in the image is a live person.
    Returns True for live, False for spoof.
    """
    # Placeholder for a liveness detection model.
    return True

def liveness_check_active(video_path):
    """
    Performs an active liveness check on a video feed.
    A real implementation would check for user actions like blinks or head turns.
    """
    # Placeholder for active liveness detection logic.
    return True

def run_face_verification(webcam_video, id_card_image):
    """
    Compares a live webcam video with an ID card image for face verification.
    """
    # Check for inputs
    if webcam_video is None or id_card_image is None:
        return "Please provide both a live video and an ID card image.", None, None

    # Step 1: Liveness Detection and Image Capture
    if not liveness_check_active(webcam_video):
        return "Active liveness check failed. This appears to be a spoof attempt.", None, None
    
    # Use the first frame from the video for face matching
    cap = cv2.VideoCapture(webcam_video)
    ret, frame = cap.read()
    if not ret:
        return "Could not capture a frame from the video.", None, None
    webcam_image_np = frame
    cap.release()
    liveness_result = "Active liveness check: Passed ✅"

    # Step 2: Face Detection & Embedding for Selfie
    faces_selfie = app.get(webcam_image_np)
    if not faces_selfie:
        return "No face detected in the video.", None, None
    face_selfie = faces_selfie[0]
    embedding_selfie = face_selfie.embedding

    # Step 3: Convert PIL Images to NumPy arrays for ID
    id_card_image_np = np.array(id_card_image)

    # Step 4: Face Detection & Embedding for ID Card Photo
    faces_id = app.get(id_card_image_np)
    if not faces_id:
        return "No face detected in the ID card image.", None, None
    face_id = faces_id[0]
    embedding_id = faces_id[0].embedding

    # Step 5: Facial Matching
    similarity_score = np.dot(embedding_selfie, embedding_id)
    
    # Define a threshold for a positive match.
    confidence_threshold = 100

    # Step 6: Verification Logic
    result_text = f"{liveness_result}\n"
    result_text += f"Similarity Score: {similarity_score:.4f}\n"

    if similarity_score >= confidence_threshold:
        result_text += "Verdict: Faces match! ✅"
        match_found = True
    else:
        result_text += "Verdict: Faces do NOT match! ❌"
        match_found = False
    
    result_text += "\n\nNote: ID text verification is not yet implemented in this demo."

    return result_text, similarity_score, match_found

# Define the Gradio interface
interface = gr.Interface(
    fn=run_face_verification,
    inputs=[
        gr.Video(label="Live Selfie (Webcam)", sources="webcam"),
        gr.Image(label="ID Card Photo", type="pil")
    ],
    outputs=[
        gr.Textbox(label="Verification Result"),
        gr.Number(label="Confidence Score"),
        gr.Checkbox(label="Match Found")
    ],
    title="Open-Source Face and ID Verification",
    description="Use your webcam for a live check to match your face with an ID photo."
)

if __name__ == "__main__":
    interface.launch()