mezanass / seg_app

full pipeline for image segmentation (dataset preprocessing, model training and model deployment) with keras and tensorflow

seg_app

The goal of this project is to train and deploy an image segmentation model, the model is an encoder-decoder that was trained on a dataset of 80 images of the ISEM 2019 class, the images were first segmented manually using VGG Image Annotator tool that produces a csv file that contains the coordinates of the various segmentation polygons, using this file a mask was created for every image, since the dataset is kinda small, during the training a data pipeline was setup to apply various transformations to the training images and masks in order to help the model generalize better. a dockerized web app was created to explore various deployment options (tensorflowjs, flask web service, ...).

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes. See deployment for notes on how to deploy the project on a live system.

Prerequisites

The main dependencies you need to install

pip install tensorflow
pip install keras
pip install Flask
pip install pymongo

Deployment

Clone the repo

git clone https://github.com/mezanass/seg_app.git
cd seg_app/deployment/flask/

run without Docker

# uncomment lines 18 and 29 from seg_app/deployment/flask/seg_app.py and comment lines 19 and 30
# install nomgodb or comment line 45 from seg_app.py
./setup.sh

run with Docker

./dockerize.sh

Built With

• Tensorflow - Training the model
• Keras - Training the model
• Flask - The web framework used
• MongoDB - The noSQL database used

Authors

• Anass Mezroui - LinkedIn
• Mustapha Bouamama - LinkedIn

License

This project is licensed under the MIT License - see the LICENSE.md file for details

Acknowledgments

• Huge thanks to the author of this notebook that was a great help