app.py

import gradio
import skimage
import scipy
import numpy
import scipy
import os

# Define the inputs that will be shown in the web GUI
inputs = [
    # gr.File(),
    gradio.Files(file_count="multiple", label="Drag and drop all images to be analyzed"),
    gradio.Radio(["Otsu", "Li"], type="value", value="Otsu", label="Select a threshold method"),
    gradio.Number(value=5, interactive=True, label="Object minimum size"),
    gradio.Number(value=20, interactive=True, label="Object maximum size"),
]

# Define the output
output = [
    gradio.File(label="Download output")
]

def example_function(image_list, threshold_method, min_size, max_size):
    """
    Example function that will threshold, label, and then filter objects
    based on size.

    This function also save the array and returns the file save path. The file
    save path will be read by gr.File and allow the user to download the
    analysed image.
    """
    # Store output filenames to be returned
    output_filelist = list()

    for image_path in image_list:
        # Load the image
        image = skimage.io.imread(image_path.name)

        # Threshold the image
        if threshold_method.casefold() == "otsu":
            th_image = image > skimage.filters.threshold_otsu(image)
        elif threshold_method.casefold() == "li":
            th_image = image > skimage.filters.threshold_li(image)
        else:
            raise NotImplementedError
        
        # Label the image
        labeled_image = skimage.measure.label(th_image)

        areas = scipy.ndimage.sum(
            numpy.ones(labeled_image.shape),
            labeled_image,
            numpy.array(list(range(0, labeled_image.max() + 1)), dtype=numpy.int32),
        )
        areas = numpy.array(areas, dtype=int)
        min_allowed_area = numpy.pi * (min_size * min_size) / 4
        max_allowed_area = numpy.pi * (max_size * max_size) / 4
        # area_image has the area of the object at every pixel within the object
        area_image = areas[labeled_image]
        labeled_image[area_image < min_allowed_area] = 0
        labeled_image[area_image > max_allowed_area] = 0

        # Construct filename from input
        base_filename = os.path.basename(image_path.name).split(".")[0]
        base_dir = os.path.dirname(image_path.name)
        output_filename = os.path.join(base_dir, base_filename + "_output.tiff")
        
        # Save file
        skimage.io.imsave(output_filename, labeled_image)

        output_filelist.append(output_filename)

    # Return the file name
    return output_filelist

demo = gradio.Interface(
    fn=example_function, 
    inputs=inputs, 
    outputs=output,
    title="Simple ID objects"
    )

if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", server_port=8000)