app.py

import gradio as gr
import cv2
import requests
import os

from ultralytics import YOLO

model = YOLO('best.pt')
image_paths = ['pothole_example.jpg', 'pothole_screenshot.png']

def show_preds_image(image):
    # Save the uploaded image temporarily
    image_path = "uploaded_image.jpg"
    cv2.imwrite(image_path, image[:, :, ::-1])  # Convert BGR to RGB and save the image
    
    image = cv2.imread(image_path)
    outputs = model.predict(source=image_path)
    results = outputs[0].cpu().numpy()
    for i, det in enumerate(results.boxes.xyxy):
        cv2.rectangle(
            image,
            (int(det[0]), int(det[1])),
            (int(det[2]), int(det[3])),
            color=(0, 0, 255),
            thickness=2,
            lineType=cv2.LINE_AA
        )
    os.remove(image_path)  # Remove the temporary image file
    return cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

inputs_image = [
    gr.inputs.Image(label="Upload Image"),
]
outputs_image = [
    gr.outputs.Image(type="numpy"),
]
interface_image = gr.Interface(
    fn=show_preds_image,
    inputs=inputs_image,
    outputs=outputs_image,
    title="Pothole detector",
    examples=[],
    cache_examples=False,
)

interface_image.launch()