siruku6/ml_sample

ML_Sample

This gif image is showing this app under development.

1. Requirement

You need to setup these and to enable docker-compose to run.

• docker
• docker-compose

Other Middlewares

• pipenv
• postgresql

*You don't have to setup these Middlewares in your OS, if you build your environment with docker and docker-compose !!

2. How to build development environment

Please refer to this doc

ml_sample's Issues

introduce easy machine learning feature

enable docker-compose on this app

Remove heavy module and model-file

refine UI

Add Text message and Link

• the bias that this model recognize the shape of mouth much, but don't recognize the shape of eyes much
  • it may be because training data don't include Japanese or any Asian faces (I heard that people in Europe or America express smile with mouth, but not with eyes).
• link to the page detect
• link to github repository including the source which made face expression detection model

Exception Handling

notify user of following things in Japanese

• camera is not usable when it is

Add easy feature

• remain displaying the most positive image on browser at another area

Other

• modify the text size of the message telling detection result

Enable user to login

The size of source is too large for deploying to heroku

-----> Building on the Heroku-20 stack
-----> Using buildpack: heroku/python
-----> Python app detected
-----> Using Python version specified in runtime.txt
 !     Python has released a security update! Please consider upgrading to python-3.9.11
       Learn More: https://devcenter.heroku.com/articles/python-runtimes
cp: cannot stat '/tmp/build_cfcd8af5/requirements.txt': No such file or directory
-----> Installing python-3.9.10
-----> Installing pip 21.3.1, setuptools 57.5.0 and wheel 0.37.0
-----> Installing dependencies with Pipenv 2020.11.15
       Installing dependencies from Pipfile.lock (69a402)...
-----> Installing SQLite3
-----> Skipping Django collectstatic since the env var DISABLE_COLLECTSTATIC is set.
-----> Discovering process types
       Procfile declares types -> web
-----> Compressing...
 !     Compiled slug size: 977.5M is too large (max is 500M).
 !     See: http://devcenter.heroku.com/articles/slug-size
 !     Push failed

possible resolutions

• https://stackoverflow.com/questions/57647989/pipenv-option-to-mimic-pip-f-option
• https://stackoverflow.com/questions/63974588/how-to-install-pytorch-with-pipenv-and-save-it-to-pipfile-and-pipfile-lock
• https://qiita.com/kimisyo/items/e73844cf1ff46e2f26c5
• https://qiita.com/shimajiroxyz/items/42bdfb412fac7d4483b7
• https://tsukimitech.com/python-ai-web-heroku-pytorch/#toc2

enable CI with github actions

draw the detection result of Face Recognition in the image displayed on browser

modify image which is captured with camera and display that image on browser

add env variables for authentication of AWS

Enable user to logout

add feature turning on camera and sending image to django server

How to use opencv to enable web-camera on local PC.

• Google Colaboratory+OpenCVでWebカメラ画像から顔検出
  This is the nearest to what I wanted at the moment.
• Sanjeevsd/yolo-with-django
  This requires that Django is running on local PC installed with camera.
• [Python][Django] StreamingHttpResponseでカメラストリーミング
  This requires that Django is running on local PC installed with camera.
• OpenCV 接続したカメラから動画を取得しよう (Python)
  This requires that Django is running on local PC installed with camera.
• USBカメラで動画撮影しよう！(on python3)
• yolo v5で自作データを学習させ、PCカメラですみっコぐらしたちをリアルタイムに物体検出してみる
  This doesn't assume taking picture with a camera on PC.
• cv2.VideoCapture(0)をしても内蔵カメラの動画が取得できない時の対処法

aa

-----> Building on the Heroku-20 stack
-----> Using buildpack: heroku/python
-----> Python app detected
-----> Using Python version specified in runtime.txt
 !     Python has released a security update! Please consider upgrading to python-3.9.11
       Learn More: https://devcenter.heroku.com/articles/python-runtimes
cp: cannot stat '/tmp/build_cfcd8af5/requirements.txt': No such file or directory
-----> Installing python-3.9.10
-----> Installing pip 21.3.1, setuptools 57.5.0 and wheel 0.37.0
-----> Installing dependencies with Pipenv 2020.11.15
       Installing dependencies from Pipfile.lock (69a402)...
-----> Installing SQLite3
-----> Skipping Django collectstatic since the env var DISABLE_COLLECTSTATIC is set.
-----> Discovering process types
       Procfile declares types -> web
-----> Compressing...
 !     Compiled slug size: 977.5M is too large (max is 500M).
 !     See: http://devcenter.heroku.com/articles/slug-size
 !     Push failed

Enable to delete customer

enable to signup

enable to run on heroku

switch the user in docker image from root to other

Now the following warning is arising while running docker-compose build.

WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv

force user to login and some fixes

Easily failed and response 503 because of running out of memory

This is the log displayed on console of heroku.

2022-05-08T03:49:57.936246+00:00 heroku[web.1]: Process running mem=737M(141.4%)
2022-05-08T03:49:57.938415+00:00 heroku[web.1]: Error R14 (Memory quota exceeded)

TODO

Show reason why app doesn't any response to user.

implement a feature to classify images

These pages seem to be helpful.

• https://blog.narito.ninja/detail/92/#_3
• https://webbibouroku.com/Blog/Article/js-file-preview#outline__2