Implementing Artificial Neural Networks with Tensorflow - Final Task: A recipe generator

A project by Antonia Hain, Anton Laukemper & Moritz Nipshagen

Preface

This project was developed in the context of the course Implementing Artificial Neural Networks with Tensorflow taught at Osnabrück University by Lukas Braun. Its goal was to test and enhance our skillset and understanding of neural networks and the tensorflow framework.

The code was developed and tested with

• python 3.6
• tensorflow 1.5 & 1.3
• MongoDB 3.6
• Firebase 03/2018

An overview

• MongoDB
  • Download & install MongoDB
  • The application expects a local server running at mongodb://localhost:27017
  • The server either needs to host a database named iannwtf and a collection recipes or the build_db.py (see below) script can crawl and build the database from scratch. This takes a while.
  • A backup of the database can be downloaded from here and restored with the mongorestore application bundled with the MongoDB server installation. The syntax is mongorestore -d iannwtf <directory_backup> or mongorestore ./ if the iannwtf folder is in the working directory. Also refer to the documentation.
• The Jupyter-Notebook file holds the actual network code.
• The backup is already processed, but if the database is created from scratch, the remap_food_ids function from the process.py script must be called beforehand.