app.py

import gradio as gr  # Import Gradio library for creating UI
import torch  # Import PyTorch for YOLOv5 model
from PIL import Image  # Import Pillow for image manipulation

# Load YOLOv5 model from Ultralytics' GitHub repository
model = torch.hub.load('ultralytics/yolov5', 'custom', path='best.pt', _verbose=False)

# Define a function to predict objects in an image
def predict(image_in_img, image_in_video):
    global model  # Use the global YOLOv5 model
    if image_in_video == None and image_in_img == None:  # If both inputs are None, raise an error
        raise gr.Error("Please upload an image.")
    if image_in_video or image_in_img:  # If either input is not None,
        image = image_in_video or image_in_img # set the image variable to the non-None input
        return model(image).render()[0]  # Use the YOLOv5 model to predict objects in the image and return the rendered output

# Define a function to toggle between webcam and file inputs
def toggle(choice):
    if choice == "webcam":  # If "webcam" is selected,
        return gr.update(visible=True, value=None), gr.update(visible=False, value=None)  # Show webcam input and hide file input
    else:  # Otherwise, if "file" is selected,
        return gr.update(visible=False, value=None), gr.update(visible=True, value=None)  # Show file input and hide webcam input

# Examples to test
ex = [["img1.jpeg"], ["img2.jpeg"], ["img3.jpeg"]]

# Create Gradio UI blocks
with gr.Blocks() as blocks:
    gr.Markdown("# CiclopeIA: Imaginando tu futuro")  # Display text in Markdown format
    gr.Markdown("## Application based on [CiclopeIA Saturdays' project](https://medium.com/saturdays-ai/ciclopeia-imaginando-tu-entorno-14dd3781a7ac)")
    gr.Markdown("### Take a photo of a € bill or make it directly with the camera and the model will recognize its value")
    with gr.Row():  # Create a row of UI elements
        with gr.Column():  # Create a column of UI elements
            # Create a radio button to choose between webcam and file inputs
            image_or_file_opt = gr.Radio(["file", "webcam"], value="file",
                                         label="How would you like to upload your image?")
            # Create an image input for a file, initially hidden
            image_in_img = gr.Image(
                source="upload", type="filepath")
            # Create an image input for the webcam
            image_in_video = gr.Image(source="webcam", visible=False, type="filepath")
            
            # Bind the toggle function to the radio button to switch between webcam and file inputs
            image_or_file_opt.change(fn=toggle, inputs=[image_or_file_opt],
                                     outputs=[image_in_video, image_in_img], queue=False)
        with gr.Column():  # Create another column of UI elements
            # Create an output image to display the predicted objects
            image_out = gr.Image()

    # Create a button to run the prediction function and display the output image
    run_btn = gr.Button("Run")
    run_btn.click(fn=predict, inputs=[
                  image_in_img, image_in_video], outputs=[image_out])

    gr.Examples(
        examples = ex,
        inputs = [image_in_img, image_in_video],
        outputs = image_out,
    )

# Launch the Gradio UI blocks
blocks.launch()