import torch
import numpy as np
import gradio as gr
from PIL import Image

# Device configuration
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

# Load YOLOv5 model
model = torch.hub.load('ultralytics/yolov5', 'yolov5s', pretrained=True).to(device)

# Set model confidence threshold
model.conf = 0.5
if device.type == 'cuda':
    model.half()

def process_frame(image):
    """Process a video frame or image and apply YOLOv5 object detection."""
    if image is None:
        return None
    
    try:
        image_pil = Image.fromarray(image)
        with torch.no_grad():
            results = model(image_pil)
        
        rendered_images = results.render()
        return np.array(rendered_images[0]) if rendered_images else image
    
    except Exception as e:
        print(f"Error processing frame: {e}")
        return image

# Create Gradio UI
with gr.Blocks(title="Real-Time Object Detection") as app:
    gr.Markdown("# Real-Time Object Detection")

    with gr.Tabs():
        # 📷 Live Webcam Tab
        with gr.TabItem("📷 Live Camera"):
            with gr.Row():
                webcam_input = gr.Video(label="Live Feed")  
                live_output = gr.Image(label="Processed Feed")
            
            webcam_input.change(process_frame, inputs=webcam_input, outputs=live_output)

        # 🖼️ Image Upload Tab (With Submit Button)
        with gr.TabItem("🖼️ Image Upload"):
            with gr.Row():
                upload_input = gr.Image(type="numpy", label="Upload Image")
                submit_button = gr.Button("Submit") 
                upload_output = gr.Image(label="Detection Result")

            submit_button.click(process_frame, inputs=upload_input, outputs=upload_output)

app.queue().launch(server_name="0.0.0.0", server_port=7860, share=False)