app.py

import numpy as np
import matplotlib.pyplot as plt
from diffusers import DiffusionPipeline

# Load the pre-trained model
pipeline = DiffusionPipeline.from_pretrained("ankush-003/retinal_fundus")
# pipeline.to("cuda")
# gradio function for generating image
def generate_image():
    image = pipeline().images[0]
    image.save("trial.png")
    img = plt.imread("trial.png")
    # Display the image (optional)
    # plt.imshow(img)
    # plt.axis("off")
    # plt.show()
    return img

# gradio interface
# import gradio as gr
# iface = gr.Interface(fn=generate_image, inputs=None, outputs=[gr.Image(label="Generated Image", type="numpy", tool='editor')],
#                 title="Image Data Generator",
#                 description="This tool generates synthetic images using the DiffusionPipeline model.",
#                 article="### Using the Image Data Generator\n\nSimply click 'Generate Image' to create a synthetic image. The generated image will be displayed below.")
# iface.launch(debug=True)

# blocks ui
import gradio as gr

def generate_multiple(num):
    images = []
    for i in range(num):
      images.append(generate_image())
    return images    

with gr.Blocks(theme=gr.themes.Soft()) as app:
    gr.Markdown("""<h1 style="text-align: center;">Synthetic Image Generator</h1>""")
    
    with gr.Tab("Generate Single Image"):
      with gr.Row():
        with gr.Column():
          gr.Markdown("""## Using the Synthetic Image Generator\n\nSimply click 'Generate Image' to create a synthetic image.\n""")
          gr.Image("./train.png",label="Training Image sample").style( rounded=True, scale=1)
          gen_button = gr.Button("Generate", variant="primary")
        gen_img = gr.Image( tool="select", type="numpy", label="Generated Image").style(height=512, width=512, rounded=True)

    with gr.Tab("Generate Multiple Images"):
      gr.Markdown(
          """
          ## Using the Synthetic Image Generator to generate multiple images
          """
      )
      gen_number = gr.Slider(2, 5, step=1.0, label="Number of Images", info="Generate multiple images")
      gen_images = gr.Gallery(label="Generated Images").style(columns=[2], rows=[2], object_fit="contain", height="auto")
      gen_m_button = gr.Button("Generate Images", variant="primary")
        
    with gr.Accordion("Read More"):
        gr.Markdown("""
        - [Images used to train the model](https://ieee-dataport.org/open-access/retinal-fundus-multi-disease-image-dataset-rfmid)
        """)

    gen_button.click(generate_image, inputs=None, outputs=gen_img)
    gen_m_button.click(generate_multiple, inputs=gen_number, outputs=gen_images)
    
app.launch(debug=True)