cancel
Showing results for 
Search instead for 
Did you mean: 

Is there a way to simulate an STM32 board's behavior without having the board

SYüce.1
Associate

Hi,

I am trying to simulate the behavior of different STM32 MCUs to see which unit meets my expectations for an object detection task.

I was wondering if there is a way to simulate their behavior on object detection problems with tflite model + validation dataset given? If so, can you help me about it?

Thanks a lot in advance!

1 ACCEPTED SOLUTION

Accepted Solutions
fauvarque.daniel
ST Employee

if you are looking for checking the accuracy of generated C code versus the original model you can do it with the validation on desktop where you can pass input and output data as an NPZ file. In X-CUBE-AI inside STM32CubeMX just click on the validate on desktop button.

If you want to have inference time numbers for your network on different STM32 boards, you can use the STM32Cube.AI Developer Cloud service (https://stm32ai-cs.st.com/). On that cloud service you can run your network on several STM32 targets that we have on our farm. In this cloud service you can also try different generation options and quantize your network (if your network is a keras network).

Regards


In order to give better visibility on the answered topics, please click on 'Accept as Solution' on the reply which solved your issue or answered your question.

View solution in original post

1 REPLY 1
fauvarque.daniel
ST Employee

if you are looking for checking the accuracy of generated C code versus the original model you can do it with the validation on desktop where you can pass input and output data as an NPZ file. In X-CUBE-AI inside STM32CubeMX just click on the validate on desktop button.

If you want to have inference time numbers for your network on different STM32 boards, you can use the STM32Cube.AI Developer Cloud service (https://stm32ai-cs.st.com/). On that cloud service you can run your network on several STM32 targets that we have on our farm. In this cloud service you can also try different generation options and quantize your network (if your network is a keras network).

Regards


In order to give better visibility on the answered topics, please click on 'Accept as Solution' on the reply which solved your issue or answered your question.