How to Build ML Web Apps using Gradio

Introduction

Gradio is a Python package that allows developers to create customizable and interactive interfaces for Machine Learning models. Deployed as a web application, Gradio allows you to build a production-ready application on top of your machine learning model and avail to Internet users using your Server IP address or domain name.

This article explains how to build machine learning web applications using Gradio. You will create a Gradio interface, integrate a working machine learning model into the web application interface, and implement additional interface features using a Vultr Cloud Server.

Prerequisites

Before you begin:

• Deploy a Ubuntu A100 Cloud GPU server.

• Using SSH, access the server

• Create a non-root user with sudo privileges. For example gradio_user

• Switch to the sudo user account

    # su gradio_user

Create the Gradio Interface

To build a new interface, install Gradio on the server, and set up your application as described in the steps below.

1. Using the Python pip package manager, install the Gradio package

    $ pip3 install gradio
   

2. Using a text editor such as Nano, create a new Python file app.py

    $ nano app.py
   

3. Add the following code to the file

    import gradio as gr
    def test(name):
        return "Hi " + name + "! Welcome to the App"
   
    interface = gr.Interface(fn=test, inputs= gr.Textbox(), outputs= gr.Textbox())
   
    interface.launch(server_name="0.0.0.0", server_port=8080) 

   Save and close the file

   Below is what the application code does:

   • import: Imports the installed Gradio package
   • def test(name): Defines a function test() with one parameter name which consists of a name given as an input to the function by the user
   • interface: Defines the function interface which includes:
     • gr.Interface: A Gradio library class used to define and configure user interfaces
     • fn = test: Passes the defined function test(). The function can consist of the machine learning model, takes inputs and produces outputs
     • inputs: Specifies the input type the user adds to the test() function. It's set to gr.Textbox() which means the input data is text
     • outputs="text": Specifies the type of output returned by the test() function. It's set to gr.Textbox() which means the output is text data
   • interface.launch(): Specifies the Gradio interface launch platforms with the following variables:
     • server_name: Specifies Gradio listening interface. When set to 0.0.0.0, Gradio accepts incoming connections from all Server IP Addresses. Set the value to 127.0.0.1 when using a reverse proxy such as Nginx
     • server_port: Set the Gradio port 8080

4. To accept incoming connections to the Gradio interface, allow the listening port 8080 through the UFW firewall

    $ sudo ufw allow 8080
   

5. Restart the firewall to save changes

    $ sudo ufw reload
   

6. Using Python, run the Gradio application file

    $ python3 app.py

   When successful, your output should look like the one below:

    Running on local URL:  http://0.0.0.0:8080
   

7. Using a web browser such as Firefox, visit your public Server IP Address on port 8080

    http://SERVER-IP:8080

   

   Verify that the Gradio web application loads successfully. Then, enter your name in the name field and click Submit to view the output. In your terminal session, press Ctrl + C to stop the application.

Example: Build an Image Restoration Web Application using Gradio

To leverage the power of your Gradio web application, build and integrate the GFPGAN model to perform image restoration tasks with the ability to upscale any low-resolution image as described below.

1. Using pip, install the latest PyTorch package

    $ pip3 install torch torchvision --index-url https://download.pytorch.org/whl/cu118 
   

2. Install GFPGAN and the model dependency packages

    $ pip3 install realesrgan gfpgan basicsr
   

3. Create a new Python application file model.py

    $ nano model.py 
   

4. Add the following code to the file

    import gradio as gr
    from gfpgan import GFPGANer
    from basicsr.archs.rrdbnet_arch import RRDBNet
    from realesrgan import RealESRGANer
    import numpy as np
    import cv2
    import requests
    
    def enhance_image(input_image):
        arch = 'clean'
        model_name = 'GFPGANv1.4'
        gfpgan_checkpoint = 'https://github.com/TencentARC/GFPGAN/releases/download/v1.3.4/GFPGANv1.4.pth'
        realersgan_checkpoint = 'https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.1/RealESRGAN_x2plus.pth'
   
        rrdbnet = RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64, num_block=23, num_grow_ch=32, scale=2)
   
        bg_upsampler = RealESRGANer(
            scale=2,
            model_path=realersgan_checkpoint,
            model=rrdbnet,
            tile=400,
            tile_pad=10,
            pre_pad=0,
            half=True
        )
   
        restorer = GFPGANer(
            model_path=gfpgan_checkpoint,
            upscale=2,
            arch=arch,
            channel_multiplier=2,
            bg_upsampler=bg_upsampler
        )
   
        input_image = input_image.astype(np.uint8)
        cropped_faces, restored_faces, restored_img = restorer.enhance(input_image)
   
        return restored_faces[0], restored_img
   
    interface = gr.Interface(
        fn=enhance_image,
        inputs=gr.Image(),
        outputs=[gr.Image(), gr.Image()],
        live=True,
        title="Face Enhancement with GFPGAN",
        description="Upload an image of a face and see it enhanced using GFPGAN. Two outputs will be displayed: restored_faces and restored_img."
    )   
   
    interface.launch(server_name="0.0.0.0", server_port=8080)

   Save and close the file

   Below is what the above code does:

   • import: Imports all the model-specific libraries
   • def enhance_image(): Stores the model checkpoints and enables the following functions:
     • restorer: Stores the image enhancement model parameters
     • bg_upsampler: Handles background enhancement model parameters
     • input_image: Handles image preprocessing
     • arch: When set to clean, it uses the clean version without the StyleGAN2-specific compiled CUDA extensions
     • model_name, gfpgan_checkpoint, realersgan_checkpoint: Consists of the URL that points to the checkpoints
     • bg_upsampler: Enables background enhancement
     • restorer: Allows face enhancement
     • input_image: Converts the input data type array to numpy.uint8
     • restorer.enhance(input_image): Stores three input types, a cropped image, restored face, and the restored image
     • return: Returns the two array variable as Gradio output
   • interface: Specifies the variable that stores the gr.Interface() function and its parameters including:
     • fn: Sets the function name Gradio needs to process. In this case, enhance_image
     • inputs: Specifies that the input type is an image.
     • outputs: Specifies that the output type is an image. Two gr.Image() functions return two images, a restored face and the restored image.
     • live: Specifies that when an input image loads, it immediately processes the output without the need to click a Submit button.
     • title: Sets the interface heading
     • description: Sets the interface description that appears below the title
   • interface.launch(server_name="0.0.0.0", server_port=8080): Specifies the compilation of the interface and sets the launch code

5. Run the Python application file

    $ python3 model.py
   

6. In a new web browser window, access the Gradio web application interface

    https://SERVER-IP:8080

   

   Verify that you can upload an input image and two restored images display in your output.

Customize the Gradio Interface

The example web application in this article uses the image Gradio type to generate the interface. To customize your application, you can implement any of the following supported Gradio input and output functions:

• Text: gr.Textbox()
• Audio: gr.Audio()
• Video: gr.Video()
• Image: gr.Image()

For example, to implement Text input and output type, apply the function to your respective interface as below:

interface = gr.Interface(
    fn=FUNCTION_NAME,  # Your Desired Function Name
    inputs=gr.Textbox(),  
    outputs=gr.Textbox()
)

To implement more than one function in a single interface, set your interface to the following

interface = gr.Interface(
    fn=FUNCTION_NAME,  # Your Desired Function Name
    inputs=[gr.Image(), gr.Audio()],
    outputs=[gr.Video(), gr.Textbox()]
)

Depending on your model requirements, you can change the inputs and outputs types. Before implementing a function, verify if it's available for either input or output. For more input and output types, visit the Gradio Interface documentation.

Conclusion

In this article, you installed Gradio and created a sample image restoration application on a Vultr Cloud GPU server. You can customize your Gradio interface with more models, and server the web application using your domain name by integrating a reverse proxy such as Nginx for public access. To implement login functionality, visit the Gradio login button documentation page.

More Information

• Gradio Quickstart Guide
• Gradio Additional Parameter Guide
• Gradio Supported Input and Output Types
• AI Face Restoration using GFPGAN on Vultr Cloud GPU
• How to Configure Apache as a Reverse Proxy with mod_proxy