cancel
Showing results for 
Search instead for 
Did you mean: 

Machine Learning method and tools for VL53L8CX and VL53L8CH

aroshani
Associate III

Hi,

I'm using P-Nucleo-VL53L8A1 multi-zone time of flight sensors for smoke detection in the open air.

I've seen STSW-img043 MZAI_EVK_v1.0.1 for calibration and recording sensor data of VL53L8CH.

I would like to know what is the best Machine learning method (ANN, CNN or ...) for training a reliable model with small footprint for STM32WLE series.

Does STMicroelectronics support ready to go example for building and training an ML model for classification of VL53L8CH data?

Which tools should be used to convert the model to a c library for embedding the ML model in the MCU?

 

Thanks.

1 ACCEPTED SOLUTION

Accepted Solutions
John E KVAM
ST Employee

I like the first video https://www.youtube.com/watch?v=-Tcz465b8So better. I did that one at CES.

Mobile Physics indeed did take that histogram data and did build that AQI software from it. 

But they did not expose how they did it. 

There is going to be some sort of marketing agreement between Mobile Physics and ST where anyone can buy the code, perhaps bundled with the sensor itself. 

We did those videos trying to entice a large volume customer into putting the sensor in their product.

Perhaps that is you?

If you are interested, go to the support page, and look for someone in the sales channel to support you. 

Then you can get information as soon as it comes out. 

- john

 


If this or any post solves your issue, please mark them as 'Accept as Solution' It really helps. And if you notice anything wrong do not hesitate to 'Report Inappropriate Content'. Someone will review it.

View solution in original post

4 REPLIES 4
John E KVAM
ST Employee

We created the VL53L8CH data especially for researchers who wanted the information, and for A/I applications. 

And the ST group that does A/I has used this sensor for some A/I applications - but I don't believe any of them used the histograms. 

But try this...

Buy this... P-Nucleo-53L8A1 evaluation kit. 

Download the VL53L8CH GUI software on your PC.

Set up the parameters and log the data you are interested in.

Edit the log file (it's an Excel spreadsheet) to add the ground truth. 

Use a bunch of those to train your model. 

I'm not an A/I guy - I've only watched them work. (This is what they did.)

From this you should be able to work out what works for you. 

You might also post this in the

STM32 MCUs Machine learning & AI

community page. 


If this or any post solves your issue, please mark them as 'Accept as Solution' It really helps. And if you notice anything wrong do not hesitate to 'Report Inappropriate Content'. Someone will review it.
aroshani
Associate III

Thank you, John

I've done all the steps that you have mentioned.

We want to use this sensor specifically for smoke detection in the air. According to the following demos, it seems it's possible to detect smoke in the air.
https://www.youtube.com/watch?v=-Tcz465b8So

https://www.youtube.com/watch?v=_1EG7z0A2sA

I would like to know if there is any ready-to-go sample code or a trained model for smoke detection provided by ST.

 

 

 

 

John E KVAM
ST Employee

I like the first video https://www.youtube.com/watch?v=-Tcz465b8So better. I did that one at CES.

Mobile Physics indeed did take that histogram data and did build that AQI software from it. 

But they did not expose how they did it. 

There is going to be some sort of marketing agreement between Mobile Physics and ST where anyone can buy the code, perhaps bundled with the sensor itself. 

We did those videos trying to entice a large volume customer into putting the sensor in their product.

Perhaps that is you?

If you are interested, go to the support page, and look for someone in the sales channel to support you. 

Then you can get information as soon as it comes out. 

- john

 


If this or any post solves your issue, please mark them as 'Accept as Solution' It really helps. And if you notice anything wrong do not hesitate to 'Report Inappropriate Content'. Someone will review it.

Thanks, John.

Exciting and great presentation.

Thanks for the update.

I'll ask the sales channel as well.