ml dev board

[Juln]

Hello all,

I am looking for a development board for ML in combination with an ESP8266.

In my vision I measure values (e.g. temperature values) and after a while I want the ML dev board to create a tinyML model like an anomaly detection. After creating the model, the dev board needs to go into deep sleep mode. Of course I need a board with minimum power consumption and maximum processing power

I am really new in this business, so let me know if you need more information to help me. I'm trying my best to familiarize myself with these things. I hope to get some help.


Thanks

Joe

 

[FlyingDutch]

Hello all,

I am looking for a development board for ML in combination with an ESP8266.

In my vision I measure values (e.g. temperature values) and after a while I want the ML dev board to create a tinyML model like an anomaly detection. After creating the model, the dev board needs to go into deep sleep mode. Of course I need a board with minimum power consumption and maximum processing power

I am really new in this business, so let me know if you need more information to help me. I'm trying my best to familiarize myself with these things. I hope to get some help.


Thanks

Joe

hello,

ESP8266 has to little CPU power for machine learning, ESP32 is better choice. I tried smaller AI models (with TensorFlow Lite runtime and Keras framework for learning ANN) on "Raspberry Pi Compute Model 4" and smaller ANNs could be run on such hardware. Without Linux OS wher one could install Python, Tensorflow, Keras it is much harder to use AI and ANNs. Another choice is using

Sipeed Maixduino AI Development Board K210 RISC-V AI+lOT ESP32​

See link:
https://www.aliexpress.com/item/330...earchweb0_0,searchweb201602_,searchweb201603_

It has onboard

• CPU: RISC-V dual-core 64-bit, using FPU400 MHz standard frequency (overclockable) neural network processor

So it has hardware acceleration for Neural Networks., abd there is many examples for this board using AI.

Best Regargs

 

[betwixt]

Well - I've got an ESP8266 here drawing less than 10uA, running from a battery and waking up every 30 seconds to measure and send a temperature over WiFi using JSON in MQTT and it is written using Arduino IDE. The temperature sensor is a DS18B20 and it also sends the battery voltage and an incrementing serial number. It is certainly achievable and the one here has had its battery topped up with a 125x60mm PV panel and been running continuously for six months.

Can't help with tinyML though.

Brian.

 

[FlyingDutch]

Well - I've got an ESP8266 here drawing less than 10uA, running from a battery and waking up every 30 seconds to measure and send a temperature over WiFi using JSON in MQTT and it is written using Arduino IDE. The temperature sensor is a DS18B20 and it also sends the battery voltage and an incrementing serial number. It is certainly achievable and the one here has had its battery topped up with a 125x60mm PV panel and been running continuously for six months.

Can't help with tinyML though.

Brian.

Hello @betwixt,

of course sending data from temperature sensor over WiFi periodically and going to sleep mode is not an issuse with esp8266. The problem is AI model for anomaly detection running on ESP8266. One can run "Tensorflow Lite" on ESP32, but ESP8266 has to little computing power for such jobs (in my opinion).

BTW: I like Wales very much - it is beautiful

Best regards