ceccode/traffic-sign-classification

Traffic sign classifier

Traffic Sign Recognition Classifier capstone project for Machine Learning Zoomcamp course: https://github.com/alexeygrigorev/mlbookcamp-code/tree/master/course-zoomcamp/12-capstone.

Introduction

The main goal is build a model to classify traffic signs with Convolutional Neural Network.

Dataset

The dataset we’ll be using to train our own custom traffic sign classifier is the German Traffic Sign Benchmark https://www.kaggle.com/valentynsichkar/traffic-signs-preprocessed.

About the dataset

The German Traffic Sign Benchmark is a multi-class, single-image classification challenge held at the International Joint Conference on Neural Networks (IJCNN) 2011. We cordially invite researchers from relevant fields to participate: The competition is designed to allow for participation without special domain knowledge. Our benchmark has the following properties:

• Single-image, multi-class classification problem
• More than 40 classes
• More than 50,000 images in total
• Large, lifelike database

Credits

• https://benchmark.ini.rub.de/

Copy of the datase

• https://drive.google.com/file/d/1-bGLNOY0z9-UwbW5iHUyIHiFzhQNK4n5/view?usp=sharing

Labels:

• label_names.csv

Sample of dataset

Notebook

• Colab notebook: https://colab.research.google.com/drive/1Z8ktFYEDljLWVolxGTC0jFrDG5ohjpCQ#scrollTo=94vMp9UzVQrW
• Copy of the notebook here: notebook.ipynb

What you'll find in the notebook

• Data preparation and cleaning
• Model creation
  • generator and training
  • model metrics
• Data augmentation
• model fitting
• Export model in h5
• Load model and evaluation
• Predict
• Covert model in tslite from h5

Local deploy

Python required. For more see: https://packaging.python.org/tutorials/installing-packages/.

Install & run

in /src dir:

pip install pipenv
pipenv install
pipenv shell
python predict.py

Test

curl --location --request POST 'http://localhost:9696/predict' \
--header 'Content-Type: application/json' \
--data-raw '{
    "url": "http://clipart-library.com/images/gieERjykT.jpg"
}'

Deploy

Docker

docker build -t traffic-sign-classifier-prediction ./src
docker run -it -p 9696:9696 traffic-sign-classifier-prediction:latest

Deploy to heroku

Heroku account needed. Install heroku cli.

Docker file

Update docker file:

#ENTRYPOINT ["gunicorn", "--bind=0.0.0.0:9696", "predict:app"] 
ENTRYPOINT ["gunicorn", "predict:app"]

login

heroku login heroku container:login

create app in heroku

heroku create your-app-name

Push docker image to Heroku

heroku container:push web -a your-app-name

Deploy container on Heroku

heroku container:release web -a your-app-name

Launch app

Test is up: https://your-app-name-xyz.herokuapp.com/heartbeat

LICENSE

LICENSE