KE Forums KE product support community

Working with a MAVN simulation model...

Discussions relating to the open source Java application MAVN (Machine Artificial Vision Network).

by Kaleb » Tue May 06, 2014 11:45 am

Create a project folder:

Before you build a new simulation, you need a folder to save your files in. Go to your Desktop and right click. Select New->Folder from the drop down options. Rename the new folder you created BigSquare by right clicking on the folder and selecting Rename.

Load a project:

To load our simulation model into MAVN 5.0, click View-> Model to get to the Model Control view. This view allows you to enter the matrices in a spreadsheet like environment.

How to Enter the Data:

Enter each matrix into the corresponding spreadsheet area starting from the cell in the upper-left-most corner. The MAVN simulation requires four input models and a fifth, the target, is optional. The models each have their own spreadsheet area and are organized according to their order of usage within the MAVN algorithm itself. A detailed explanation of what each model does in the MAVN simulation can be found in the tutorials section. After you have entered your model data into the spreadsheet, be sure to press the Set Model button so the simulation can use it.

The MAVN 5.0 Input Models:

The Input Model Views within MAVN 5.0 are rendered with spreadsheets. Each input for the model for the simulation has its own spreadsheet. MAVN 5.0 simulations require at least four different input models: W0, W1, W2 and Theta. The spreadsheets act like any generic spreadsheet software, providing copy/paste, delete and undo/redo functionality.


Load a Simulation Model:

The easiest and fastest way to learn MAVN 5.0 is to just load an example model. To load an example simulation, on the menu bar click File-> Open Simulation. Alternatively, you can click the Open Simulation button in the control bar.


File Chooser:

Use the File Chooser dialog box to navigate to the MAVN 5.0 folder, open it and then open the Examples folder. Select (just select it so the folder is highlight, don’t double-click it to enter the folder) the DiamondHorseshoe folder and click the Open button.


Supported file types:

MAVN 5.0 can import simulations from Comma Delimited file formats such as .CSV and the Microsoft file formats .XLS and .XLSX. It can also save input models to these same file formats. Any of the MAVN 5.0 input model files can be edited with Microsoft Excel, Open Office, Lotus Notes or any other spreadsheet like software that supports Comma Delimited or Microsoft Spreadsheet file formats. Once you have loaded the required input models into the program, MAVN 5.0 will automatically detect if the model is valid and then render the network so you can begin your simulations.

Run a Simulation:

MAVN was designed to deliver a high level understanding of how the simulation works before trying to understand the details of the underlying model. Before getting too bogged down in all the details of what each part of the model does or how the simulation works, let’s just run a simulation. The default simulation for MAVN is the Monte Carlo Pattern Recognition Simulation.


When the Simulation Runs:

The network nodes will begin to flash black and green as the simulation runs. A black node means the node is not active, and a green node means the node is active. At the end of the simulation, the results are displayed. The black dots on the image were points that missed shapes. The red dots are points that hit shapes. The ratio of the number of darts that hit each shape versus the total number of darts fired is displayed in the results panel. These are the outputs of the simulation that can be fed into other networks to determine if it is a pattern that can be recognized.


Modify the Simulation Properties:

You can change the type of simulation that will be used to run the simulation from the input models with the Simulation Properties View or with the MAVN Control Bar.

To quickly change simulation types, simply click on the Diagnostic Simulation, Monte Carlo Simulation, or Pixel Simulation buttons in the MAVN Control Bar, respectively.

Change the Simulation Properties:

To change the Simulations Properties, click the Simulation Properties button on the far left of the Properties Control Bar to open the Simulation Properties View.

The Simulation Properties View will allow you to define the type of simulation you are going to run. The properties let you control what type of simulation is run, how many points should be used in the image, the random number generator, and the random number seed. Between these options, there are huge number of different simulation that can be run using different technique to optimize speed and image resolution.

How to Save the Model:

After you have entered the model matricies into the spreadsheets, click the Set Matrix buttons to load each matrix into the simulation. Once you have done this, you can save the file into the BigSqaure folder you created (use whatever folder name you want).

Save the Model:

To save the model to a file, click the Save Model button for the spreadsheet you want to save. The MAVN application uses a very specific naming convention so you can store all the models related to a single simulation in one folder. Each model is stored in one file, either a .CSV, .XLS, or .XLSX file. Storing them together in one folder allows the application to quickly load simulations with one action.

Name the model files for each simulation as following in one folder/directory:

You must name files using the following conventions. Then you will be able to quickly load the entire simulation from the folder with one mouse click the next time you need it.

Use the following naming conventions for your models:

Target Matrix = “Target”
Theta Matrix = “Theta”
W0 Weight Matrix = “W0”
W1 Weight Matrix = “W1”
W2 Weight Matrix = “W2”
Site Admin
Posts: 11
Joined: Tue Apr 01, 2014 2:26 pm

by Sefirotara » Sun Mar 29, 2015 11:51 pm

I am grateful to suppress That have come into this website.
Posts: 2
Joined: Sun Mar 29, 2015 10:28 pm

Return to MAVN (Machine Artificial Vision Network)