import cv2
import numpy as np
import torch
import streamlit as st
from ultralytics import YOLO
from camera_input_live import camera_input_live

# Load YOLO fire detection model
model_path = "last.pt"
if not torch.cuda.is_available():
    device = "cpu"
else:
    device = "cuda"

model = YOLO(model_path)
model.to(device)

# Streamlit app title
st.title("Live Fire Detection with Camera")
st.subheader("Hold the camera towards potential fire sources to detect in real-time.")

# Capture live camera input
image = camera_input_live()

if image is not None:
    # Convert the image to OpenCV format
    bytes_data = image.getvalue()
    cv2_img = cv2.imdecode(np.frombuffer(bytes_data, np.uint8), cv2.IMREAD_COLOR)

    # Perform fire detection
    results = model(cv2_img, device=device)

    # Draw bounding boxes for detected fires
    for result in results:
        boxes = result.boxes
        for box in boxes:
            b = box.xyxy[0].cpu().numpy().astype(int)
            c = int(box.cls[0])
            label = f'Fire {box.conf[0]:.2f}'
            cv2.rectangle(cv2_img, (b[0], b[1]), (b[2], b[3]), (0, 0, 255), 3)
            cv2.putText(cv2_img, label, (b[0], b[1] - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.9, (0, 0, 255), 2)

    # Display the annotated image
    st.image(cv2_img, channels="BGR", caption="Detected Fire", use_container_width=True)