gagahgk
September 17, 2020, 1:48pm
1
hello, sorry if this has been asked before
i’m trying to use libraries from tensorflow followed by this article
https://towardsdatascience.com/tensorflow-meet-the-esp32-3ac36d7f32c7
there’s some slight modification to the steps since the new tensorflow lite is no longer an experimental
i moved all the libs from the compilation to the “lib” folder in the project you can see the structure below
and after doing that i tried to include the libs but it doesn’t recognized by the IDE
i tried to compile it anyway but i got error saying “no such file or dirrectory”
did i placed the libs incorrectly ?
i changed the lib_ldf_mode = deep still didnt fix it
English is not my main language sorry if there’s some part you didn’t understand, really apreciate help from this community thankyou
Their sample project is up at GitHub - wezleysherman/ESP32-TensorFlow-Lite-Sample: Sample project for deploying TensorFlow Lite models on the ESP32 using Platformio and you can look at the folder structure there. It’s different to yours since it has lib/tfmicro/tensorflow/lite
in it, but you’re doing lib/tensorflow/lite/micro
.
1 Like
gagahgk
September 17, 2020, 3:12pm
3
i make some adjustment according to the GitHub Page turns out the problem is because i didn’t include the
libs_deps = tfmicro
in platformio.ini
the dependency is now fixed but i got another problem, i think it has to do with the toolchain the error
the code looks like this
the problem
and the compile log
tried to change the “include” with
#include <tgmath.h>
i get more errors instead
gagahgk:
So have you just tried enabling the other codepath and use std::max
instead of std::fmax
. Because that at least auto-completes for me in the ESP32 environment.
Adding something like || defined(ARDUINO)
would suffice in the #if defined
check expresson. Or ARDUINO_ARCH_ESP32
if you want to be precise and limiting to ESP32-Arduino.
Same for fmin()
.
1 Like
gagahgk
September 17, 2020, 3:38pm
5
adding the defined(arduino) fixed the problem for me thankyou