import gradio as gr
import torch
import cv2
from ultralytics import YOLO

# Load YOLO models
def safe_load_yolo_model(path):
    torch.serialization.add_safe_globals([torch, 'ultralytics.nn.tasks.DetectionModel'])
    return YOLO(path)

# Load the models
model_yolo11 = safe_load_yolo_model('./data/yolo11n.pt')
model_best = safe_load_yolo_model('./data/best.pt')

# Class names for YOLO model (replace with actual class names used in your YOLO model)
yolo_classes = ['person', 'bicycle', 'car', 'motorcycle', 'airplane', 'bus', 'train', 'truck', 'boat', 'traffic light', 'fire hydrant', 'stop sign', 'parking meter', 'bench', 'bird', 'cat', 'dog', 'horse', 'sheep', 'cow', 'elephant', 'bear', 'zebra', 'giraffe']

# Class names for best.pt model (assumed classes for crack and pothole)
best_classes = ['Crack', 'Pothole']

def process_video(video):
    # Open the video using OpenCV
    cap = cv2.VideoCapture(video)
    fps = cap.get(cv2.CAP_PROP_FPS)
    frame_width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
    frame_height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))

    # Create VideoWriter to save output video
    fourcc = cv2.VideoWriter_fourcc(*'mp4v')  # Codec for .mp4
    out = cv2.VideoWriter('output_video.mp4', fourcc, fps, (frame_width, frame_height))

    while cap.isOpened():
        ret, frame = cap.read()
        if not ret:
            break
        
        # Detect with YOLOv11 (general object detection model)
        results_yolo11 = model_yolo11(frame)
        # Detect with best.pt (specialized model for cracks and potholes)
        results_best = model_best(frame)

        # Draw bounding boxes and labels for YOLOv11 (General Object Detection)
        for result in results_yolo11:
            boxes = result.boxes
            for box in boxes:
                x1, y1, x2, y2 = map(int, box.xyxy[0].tolist())
                class_id = int(box.cls[0])  # Class index for YOLO
                label = f"YOLO: {yolo_classes[class_id]} - {box.conf[0]:.2f}"  # Map class_id to class name
                cv2.rectangle(frame, (x1, y1), (x2, y2), (0, 255, 0), 2)
                cv2.putText(frame, label, (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.9, (0, 255, 0), 2)

        # Draw bounding boxes and labels for best.pt (Crack and Pothole detection)
        for result in results_best:
            boxes = result.boxes
            for box in boxes:
                x1, y1, x2, y2 = map(int, box.xyxy[0].tolist())
                class_id = int(box.cls[0])  # Class index for best.pt
                label = f"Best: {best_classes[class_id]} - {box.conf[0]:.2f}"  # Map class_id to specific labels
                cv2.rectangle(frame, (x1, y1), (x2, y2), (255, 0, 0), 2)
                cv2.putText(frame, label, (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.9, (255, 0, 0), 2)

        # Write the processed frame to the output video
        out.write(frame)

    cap.release()
    out.release()

    return 'output_video.mp4'

# Gradio interface
iface = gr.Interface(fn=process_video, inputs=gr.Video(), outputs=gr.Video(), live=True)

# Launch the app
iface.launch()