Deploy Vision Model on PSOC™ Edge boards
In this section, we cover information on how to deploy the vision code example on the PSoC™ Edge using ModusToolbox™.
How to deploy the model onto the board?
After the code is generated, follow the step-by-step instructions to deploy the model onto the board.
Optional: Placing the generated source in your ModusToolbox™ Application
If you have not set the Output folder to generate code in the required ModusToolbox™ project, follow these steps, otherwise proceed to the Building and running the application section.
-
Copy the output directory from the path defined in Output folder. The default folder name is mtb_ml_gen.
-
Paste the folder within the proj_cm55 folder of your ModusToolbox™ project.
Building and running the application
- Right-click the project and select Build Project or click Build Application in the Quick panel to build the project.
For detailed instructions on how to run and verify the application live on your PSOC™ Edge kit, read the README.md in the ModusToolbox™ application or here (opens in a new tab).
Updating the classes
The ModusToolbox™ vision template application contains a hard-coded list of class strings. This is used to map the output of the Object Detection model into readable strings. These strings are used to display object names on the screen attached to your PSOC™ Edge kit. It is also used for printing information to the terminal. If you have trained or downloaded a model that detects different classes than the template application, you need to update the list accordingly in 'proj_cm55/source/model_info.h'.
If you have trained the model in Studio, make sure the order of the classes in the class list string matches the order in your DEEPCRAFT™ Studio project. To see the order of the classes in Studio, open the project file(.improj) and select the Data tab.
Validating the generated source
This is an optional step but if you want to validate the model using the ML Configurator. Refer to ML Configurator user guide (opens in a new tab).