ESP32 Forum

Hi, I need to detect objects from the captured frames of the esp-eye. In order to do that, I have trained a tensorflow model following this tutorial https://tensorflow-object-detection-api ... ining.html and as I can see here viewtopic.php?p=47138#p47138 it has been explained that there is a way to convert a tensorflow model to be read by the esp-who library (dl_lib.h). But I cannot understand how to do such conversion from a tensorflow model (.pb file). Can please anyone guide me through this operation with some more detailed advices? Thank you in advance.

hey,
so i'm still a bit lost on this as well (im very new to deep learning)
from my understanding you need to train the model using whatever you like (tensorflow, pytorch, whatever..)
to get your inference graph...once you have that its

when you build the model in C for the ESP you can use a lot of functions in the ESP-WHO liibrary
mainly:
image_util.h
fb_gfx.h
fd_forward.h
fr_forward.h
mtmn.h
dl_lib.h
dl_lib_matrix3dq.h
they help get your data into the right format and perform operations you need
(look at the header files for names/explanations of the functions available)
finally, once all thats done youll want to make your inference
I'm sorry i dont have code yet, i'm still working it all out myself

Hey, as an update I think i've figured out how to get custom object detection working
( the code has a few bugs that I'm working out, but the model does run)
code is found here:
https://github.com/caleb221/ESP32-Leaf