View the whole notebook: https://github.com/mattclifford1/IQM-Vis/tree/main/dev_resources/docs/notebooks/Tutorial_2-Customisation.ipynb

---

Tutorial 2: Simple Customisation

In this notebook we will get to grips with how to peice together a simple IQM-Vis UI. We will go through the 3 basic aspects of customisation: - Images - Metrics - Transformations

Let’s start by importing the package.

import IQM_Vis

Images

To include custom images we define their filepaths. For now we will use the example images provided with IQM-Vis, but feel free to change the file paths with your own local image files.

image1 = IQM_Vis.examples.images.IMAGE1
image2 = IQM_Vis.examples.images.IMAGE2

images = [image1, image2]
print(f'Images files: {images}')

Images files: ['/home/matt/projects/IQM-VIS/IQM_Vis/examples/images/waves1.jpeg', '/home/matt/projects/IQM-VIS/IQM_Vis/examples/images/waves2.jpeg']

Metrics

To include some custom image quality metrics we store them in a dictionary. These return a scalar value when comparing two images. For now we will use some metrics provided by IQM-Vis. For a list of all metrics provided by IQM-Vis see the documentation (https://mattclifford1.github.io/IQM-Vis/IQM_Vis.metrics.html). See later tutorials for how to define your own custom metrics.

MAE = IQM_Vis.IQMs.MAE()
MSE = IQM_Vis.IQMs.MSE()
SSIM = IQM_Vis.IQMs.SSIM()

metrics = {'MAE': MAE,
           'MSE': MSE,
           '1-SSIM': SSIM}

Metric Images

Adding image quality metrics that return an image instead of a scalar value is a similar process to the metrics before. Defining these is optional but can be useful for qualitive spacial analysis.

MSE_image = IQM_Vis.IQMs.MSE(return_image=True)
SSIM_image = IQM_Vis.IQMs.SSIM(return_image=True)

metric_images = {'MSE': MSE_image,
                 '1-SSIM': SSIM_image}

Transformations

Defining image transformation/distortions requires a little more information as we also need to provide the range of values that we want the transformations to operate over.

rotation = IQM_Vis.transforms.rotation
blur = IQM_Vis.transforms.blur
brightness = IQM_Vis.transforms.brightness
jpeg_compression = IQM_Vis.transforms.jpeg_compression

transformations = {
    'rotation':  {'min':-180, 'max':180, 'function':rotation},                # normal input
    'blur':      {'min':1,    'max':41,  'function':blur, 'normalise':'odd'}, # only odd ints since it's a kernel
    'brightness':{'min':-1.0, 'max':1.0, 'function':brightness},              # float values
    'jpg comp.': {'min':1,    'max':100, 'function':jpeg_compression, 'init_value':100}, # non zero inital value
               }

Putting it all together

We need to pass everything to the UI maker

IQM_Vis.make_UI(transformations=transformations,
                image_list=images,
                metrics=metrics,
                metric_images=metric_images)

You will now get the desired UI with the specified transformations, metrics and images