STEVAL-STWINBX1 – “Automatic build and run failed” during Validation on Target

[dzf]

Dear ST Technical Support Team,

I am currently developing a voice recognition application on the STEVAL-STWINBX1. I trained a classification model using the notebook:

Utilities\AI_Resources\TrainingScripts\UltrasoundClassification\UltrasoundClassification.ipynb

In the notebook/off-target evaluation, the model achieves >90% accuracy, but after deploying to the MCU the recognition accuracy becomes very low.

To investigate, I used STM32CubeMX → Validation on Target. However, the validation process fails with the message: “Automatic build and run failed” (Automatic Validation on the target)

Thank you for your help. I look forward to your guidance on resolving the validation failure and improving on-MCU accuracy.

Best regards,
dzf

[Andrew Neil]

---

@dzf wrote:

Dear ST Technical Support Team,

---

You realise this is a Community forum - most contributors are not ST Staff ?

If you want direct support from ST, open a Support Case at https://ols.st.com/s/ 

You could also contact you local FAE: https://www.st.com/content/st_com/en/contact-us.html

[Julian E. · Accepted solution]

Hi @dzf,

I have no knowledge about this notebook.

I would say that it is pretty deprecated.

The main ways to do AI for STM32 is now to use the tools that you can find here:

Edge AI solutions overview - STMicroelectronics - STM32 AI

In your case, I would suggest to look at the ST Model Zoo Services and the audio use cases.

Have a good day,

Julian