View source for AI:How to use transfer learning to perform image classification on STM32

=== Model generation for STM32 ===
{{Warning|Warning, the latest version of FP-AI-VISION1 v3.1.0 is based on STM32Cube.AI 7.1.0. Make sure your are using this version.}}

First convert the <code>tflite</code> model to be used on a STM32 platform.

There are two ways to achieve this goal, one by using STM32CubeMX and generating an entire project (recommended at first), the other by using the command line interface to generate only the required files.

==== Project generation with STM32CubeMX (Option 1) ====

Start STM32CubeMX and create a new project. In this example, the STM32H747I-DISCO board is used. Select the board in the <code>Board Selector</code> and start the project.

A new window opens showing the board configuration. Click <code>Software Packs -&gt; Select Components</code>.

<div class="res-img">
[[File:CubeMXScreenShot.png|center|Adding components to STM32CubeMX]]
</div>

Select the components for the Arm<sup>&reg;</sup> Cortex<sup>&reg;</sup>-M7 core under STMicroelectronics X-CUBE-AI, check the core box, select <code>Application</code> as application, then click Ok.

<div class="res-img">
[[File:CubeMXScreenShot2.png|center|Components selector]]
</div>

The stm32-ai library will be download if it has not been downloaded yet. When it is complete, click Ok.

Go to <code>Software Packs (on the left) -&gt; STMicroelectronics X-CUBE-AI</code>. A pop up window opens to configure peripherals and clock.Click Yes.

Click <code>Add network</code>, call your network <code>network</code>, select TFLite from the scrollable menu, and browse to the last section to find the generated model.

Once the model is selected, click <code>Analyze</code>. This provides a quick overview of the model memory footprint. Click the <code>Show graph</code> button for more in-depth information on the model topology and RAM memory usage layer by layer including activation buffers.

<div class="res-img">
[[File:Graph1.png|center|Memory Usage windows]]
</div>

To customize how the model is stored in memory, click the gear icon (<code>Advanced Settings</code>) and specify if you want to use external Flash memory and/or RAM. Check the <code>Use activation buffer for input buffer</code> to make them share the same buffer in order to reduce memory usage.

<div class="res-img">
[[File:CubeMXScreenShot3.png|950px|center|X-Cube-AI model loading]]
</div>

Click <code>Analyze</code> again. You can see that the RAM occupation has been significantly reduced  (refer to the memory graph for more details).

<div class="res-img">
[[File:Graph2.png|center|Memory Usage window after enabling input buffer space sharing]]
</div>

Check that the model works properly by validating it either on the PC or on the target.

To generate the code for STM32CubeIDE (or Arm<sup>&reg;</sup> KEIL<sup>&reg;</sup> or IAR SYSTEMS<sup>&reg;</sup>) go to <code>Project manager</code>, specify a project name and location. Select STM32CubeIDE as toolchain, then click <code>GENERATE CODE</code> on the top right corner.

<div class="res-img">
[[File:CubeMXScreenShot4.png|950px|center|Project manager window]]
</div>

This generates the necessary code for the whole project. In the present example, only the files related to the neural network are required.

They are located under <code>YourProjectName\CM7\X-CUBE-AI\App\</code> and are called <code>network.c</code>,<code>network.h</code> ,<code>network_data.c</code>, <code>network_data.h</code> and <code>network_config.h</code>. They represent the model and its weight.

==== Command line interface (Option 2) ====

The other way to generate the necessary files is to use the command line interface. The main advantage is to produce only the required files.

The documentation related to X CUBE AI is located under <code>C:\Users\&lt;USERNAME&gt;\STM32Cube\Repository\Packs\STMicroelectronics\X-CUBE-AI\7.1.0\Documentation/</code>.

The CLI is called <code>stm32ai.exe</code> for Windows. It is located under <code>C:\Users\&lt;USERNAME&gt;\STM32Cube\Repository\Packs\STMicroelectronics\X-CUBE-AI\7.1.0\Utilities\windows</code> and <code>$HOME/STM32Cube/Repository/Packs/STMicroelectronics/X-CUBE-AI/7.1.0/Utilities/</code> for UNIX<sup>&reg;</sup>-based systems.

If there is no <code>utilities</code> folder, create a CUBE MX project and add Core AI in the Software Package (see section Project generation). This  downloads all the necessary files.

For ease of usage, add the folder to your Windows path:
 {{PC$}} set CUBE_FW_DIR=C:\Users\&lt;USERNAME&gt;\STM32Cube\Repository
 {{PC$}} set X_CUBE_AI_DIR=%CUBE_FW_DIR%\Packs\STMicroelectronics\X-CUBE-AI\7.1.0
 {{PC$}} set PATH=%X_CUBE_AI_DIR%\Utilities\windows;%PATH%

Once added, go to the folder where you wish to generate the network files in command line (using PowerShell or similar) and run the following command:

 {{PC$}} stm32ai.exe generate -m .\model_quant.tflite --allocate-inputs # (change the path to the model if needed).

This creates all the necessary files under the <code>stm32ai_output</code> folder.