User Manual
`# User requirements
The requirements that are needed to use the application are:
-
A computer available. This is a desktop application.
-
MATLAB Runtime 2023b, available at https://es.mathworks.com/products/compiler/matlab-runtime.html.
The only program required to run this project is MATLAB Runtime, specifically the version corresponding to MATLAB 2023b. To download it, go to the official MathWorks website . Once downloaded and installed, a MATLAB account is not required for use. MATLAB Runtime allows you to run applications compiled with MATLAB Compiler without needing a full MATLAB license.
There is no additional installation process apart from the dependencies mentioned above, as the application is self-contained. This means that all necessary files and libraries are already included in the release folder. Thanks to this, installation and usage are considerably simplified, allowing even users with no experience in MATLAB to have a pleasant experience.
The first step will be to go to the GitHub repository and download the latest release latest release:
Release of the InvIPM application on GitHub, including the version, a brief description of the changes and new features compared to the previous one, and a zip with the binaries to run on Linux and Windows. In this section, it is very important to download the version corresponding to the operating system used.
Once the zip file has been downloaded, we will have to unzip it. After that, we will have a series of files and folders in the desired directory. These are:
-
/InvIPM.exe: This is the application executable.
-
/readme.txt: This is a help document which tells us the necessary requirements as well as how to run the application.
-
/data: This directory stores the example images that the application itself brings, these can be deleted by the user if desired.
-
/cache: This directory stores the images saved in the cache.
-
/results: This is the directory that is offered to the user by default when he wants to save the results of an execution.
-
/splash.png: This is the image that appears as soon as the application is opened.
Section dedicated to writing a simple manual so that the user understands how to use the application.
As soon as you enter the application, you will be presented with a welcome image that explains in general terms the purpose of the application, as well as showing the author and tutors of the project. The content of this screen is shown in figure shows the InvIPM welcome window appearance.
At the top right part of this screen, there is a language selector between Spanish and English, with their corresponding flags. This makes it more self-explanatory, allowing the user to identify it without having to read the text. This is especially useful when the application is in a language other than that of the user.
In this section, the user can load, execute and view the results of the execution.
Figure shows the InvIPM algorithms exploration window appearance.
Using the ‘Load metal piece image’, the user will be able to select an image provided by the system or your own on which the algorithms will be applied later.
If the image is not provided by the system, the button ‘Upload groundtruth image (optional)’ will be enabled so that you can provide your own. This image, as indicated in the text, is not necessary for the execution except for obtaining the percentage of success.
The different types of illumination invariant methods are provided:
-
Álvarez.
-
Maddern.
-
Krajnık.
-
Upcroft.
-
PCA.
They are ordered so that the first one is the one that offers the best results according to the tests performed. On the other hand, PCA, although it shows a clear improvement in many situations, is a more classical approach that numerically does not achieve results as good as other methods, specifically Alvarez.
The following types of clustering methods are given:
-
K-Means.
-
Fuzzy C-Means.
-
Gaussian Mixtures.
-
HMRF_EM (it uses spatial information).
These are ordered so that the first is the least computationally expensive, so it will offer better execution times. The last, on the other hand, has longer execution times, since it is an algorithm that not only considers the color of a particular pixel, but also that of the surrounding pixels.
Using this text box, the user can indicate the number of centers to be used in the clustering algorithms. By default, this number will be 2.
The number of centers must meet the following conditions:
-
It has to be a positive number.
-
It has to be higher or equal to 2.
-
It has to be lower or equal to 10.
This is because at least two centers are needed to separate the part from the background. The maximum is established because the greater the number of centers, the greater the computational complexity, which would considerably lengthen the execution times.
Pressing the ‘Run’ button will start the execution of the different chosen algorithms on the selected image. During the execution, feedback will be provided on the percentage of progress and the stage that is currently being calculated.
After finishing the execution, four images will be displayed on the right side, these will be:
-
The original image (top left).
-
The original segmented image (top right).
-
The invariant image (bottom left).
-
The segmented invariant image (bottom right).
In the case where an image provided by the application was used or the user provided a reference image (groundtruth), the title of both the original segmented image and the segmented invariant image will show the percentage of success. In this way, the user will be able to compare the results not only visually, but also numerically.
This section presents a table in which all the executions carried out in this session and in previous sessions appear in list form. The content of the columns in this table is as follows:
-
File name: This is the name of the file that the user selected.
-
Date: This is the date of the execution. This is in DD-MMM-YYY HH-MM-SS format.
-
Invariant algorithm: Displays the name of the invariant algorithm applied to the image. If no invariant algorithm has been applied, it displays ‘Not applied’.
-
Clustering algorithm: Displays the name of the clustering algorithm applied to the image. If no clustering algorithm has been applied, it displays ‘Not applied’.
-
Number of centers: Displays the number of centers used in the clustering algorithm applied to the image. If no clustering algorithm has been applied, it displays ‘-’.
-
Groundtruth image: indicates with ‘Yes’ or ‘No’ whether the image execution had an associated image or the user had provided their own.
-
Quality measure: indicates the success rate achieved when comparing the image corresponding to that row with its ground truth image. If there is no ground truth image, this cell will be blank.
-
File: shows a checkbox which, when clicked, opens the image corresponding to that row in a new window. This new window has a series of options apart from the display, allowing the user to save or print it, among other options.
The user can also, if desired, sort the table by any column.
Figure shows how the InvIMP Exploration historical summary section looks like
Apart from the table, at the bottom there are two buttons, one to view the images and another to clear the contents of the cache.
Clicking this button will open a new window where you can view and select all the images that are stored in the cache.
When you click this button, a pop-up window will appear asking if you really want to permanently delete the contents of the cache. If you click the ‘Delete’ button, the contents of the cache will be deleted and the table will be updated, since these images will no longer be available. After this, it will indicate that the cache has been deleted successfully.
In this section, Figure fig:app_ayuda shows how you can obtain information about how it works, the illumination invariant algorithms, the clustering algorithms, the cache memory and the accuracy. Since the documentation itself is implemented within the application, it is very convenient to search for information on any topic of interest.
Besides, the documentation language corresponds to the one selected in the welcome part.
Figure shows how the InvIMP Help section looks like.
