[PYTHON] Image recognition with API from zero knowledge using AutoML Vision

Image recognition using the cloud can be done normally if you follow the preceding article, so you may not write much

I'll write an article I did in Edge instead

• This is also being written Image recognition of garbage with Edge (Raspberry Pi) from zero knowledge using AutoML Vsion and TPU When I did various things, the environment was broken and there were many difficulties, so it would be nice if the process could be summarized.

About the author

I have touched C, C ++, JAVA, python, chainer, etc. Totally ignorant about Raspberry Pi, GCP, API Since I usually use windows, I start by building an Ubuntu environment

environment

Ubuntu 18.04.2 LTS (Nopa) Raspbian（Raspberry Pi 3 Model B+） Coral accelerator EDGE TPU

things to do

Image recognition of garbage through the cloud using API of GCP and AutoML Vision @Ubuntu Image recognition of local garbage with Raspberry Pi using Edge function of AutoML Vision @Rasbian Construction of a garbage sorting system including hardware (I will not post anything other than image recognition to Qiita) (like an automatic sorting trash can)

What i did

Build an environment where you can run AutoML from a terminal using gcloud Model learning, etc. (If you connect GCP, you can do it with a mouse) Run AutoML from python Real-time execution of AutoML on Ubuntu + USB camera Create and execute an image data collection program with Ubuntu + USB camera Raspberry Pi + USB camera to create and execute an image data collection program

Reference (copy and paste source)

Please forgive me for having a lot of copies as it is almost my own memorandum Install pyenv and pyenv-virtualenv

imwrite https://note.nkmk.me/python-opencv-imread-imwrite/

AutoML API https://qiita.com/pyru89kwmr/items/c723bb4456443bcaafa2

Make Raspberry Pi a jupyter server http://irukanobox.blogspot.com/2018/02/raspberry-pijupyter-notebook.html https://qiita.com/shimaken/items/b411de87b00c051e6697

Turn on SSH for Raspberry Pi https://s2jp.com/2014/01/raspberry-pi-ssh/

SSH connection to Raspberry Pi on windows10 https://www.onebizlife.com/windows10-ssh-client-1980

memo install opencv I tried other than this, but I got an error version 3.1.0 at anaconda

$ conda install -c https://conda.binstar.org/menpo opencv3

An error will occur if the time is off https://qiita.com/yuji38kwmt/items/b3fe316f6c000235aab5

The initial password for Raspberry Pi is raspberry Raspberry Pi SSH valid command

sudo raspi-config

From

I get an error with wget

ERROR: The certificate of ‘dl.google.com’ is not trusted.

https://blog.dreamhive.co.jp/mkoba/?p=5741

autoML on edge https://www.apps-gcp.com/automl-vision-edge/ r

Install pyenv

It seems that the google cloud module required for AutoML only works on python2 series, so first insert pyenv It ’s more accurate to look here Install pyenv and pyenv-virtualenv

Install pyenv

git clone https://github.com/pyenv/pyenv.git ~/.pyenv

echo 'export PYENV_ROOT="$HOME/.pyenv"' >> ~/.profile
echo 'export PATH="$PYENV_ROOT/bin:$PATH"' >> ~/.profile
echo -e 'if command -v pyenv 1>/dev/null 2>&1; then\n  eval "$(pyenv init -)"\n

source ~/.profile

like this

pyenv install 3.6.3

Even if you do, an error will occur.

sudo groupadd developer
sudo gpasswd --add USERNAME developer
sudo visudo
Defaults exempt_group="developer"

python installation

pyenv install 3.6.3

Also install pyenv-virtualenv

//github.com/pyenv/pyenv-virtualenv.git $(pyenv root)/plugins/pyenv-virtualenv
echo 'eval "$(pyenv virtualenv-init -)"' >> ~/.profile
source ~/.profile
pyenv virtualenv 3.6.3 myenv

I will check it

pyenv versions
* system (set by /home/user/.pyenv/version)
  3.6.3
  3.6.3/envs/myenv
  myenv

If python -V still doesn't change, I could use this command

eval "$(pyenv init -)"

https://akamist.com/blog/archives/2610

Try the AutoML Vison API

I was able to go straight to this street https://qiita.com/pyru89kwmr/items/c723bb4456443bcaafa2