ankaiser4711 / test

erster test

based on: https://eloquentarduino.github.io/2021/06/howto-load-tensorflow-lite-tinyml-model-from-internet-on-arduino/

SDCard path not tested!

board: esp32 wroover

possible ways to load a model

• als C/C++ headerfile
• load from sdcard
• load from spiffs
• load via http

change way to load model within tflitearduino.ino:

// !!! UNCOMMENT ONLY 1 LINE !!!

• //#define MODEL_INCLUDE_TYPE_CPP
• //#define MODEL_INCLUDE_TYPE_SDCARD
• #define MODEL_INCLUDE_TYPE_SPIFFS
• //#define MODEL_INCLUDE_TYPE_HTTP

Benchmark:

dump cpp model - 21280

0x1C 0x00 0x00 0x00 0x54 0x46 0x4C 0x33 0x14 0x00 ... 0x00 0x00 0x00 0x04 0x00 0x04 0x00 0x04 0xFF 0xFF

• It took 1971 micros to run inference
• It took 723 micros to run inference
• It took 725 micros to run inference
• It took 728 micros to run inference
• It took 726 micros to run inference
• It took 714 micros to run inference

##############

dump http model - 21280

0x1C 0x00 0x00 0x00 0x54 0x46 0x4C 0x33 0x14 0x00 ... 0x00 0x00 0x00 0x04 0x00 0x04 0x00 0x04 0xFF 0xFF

• It took 561 micros to run inference
• It took 325 micros to run inference
• It took 316 micros to run inference
• It took 317 micros to run inference
• It took 316 micros to run inference
• It took 324 micros to run inference

##############

dump spiffs model - 21280

0x1C 0x00 0x00 0x00 0x54 0x46 0x4C 0x33 0x14 0x00 ... 0x00 0x00 0x00 0x04 0x00 0x04 0x00 0x04 0xFF 0xFF

• It took 563 micros to run inference
• It took 312 micros to run inference
• It took 319 micros to run inference
• It took 312 micros to run inference
• It took 312 micros to run inference
• It took 321 micros to run inference