/deep_ingredient_recognition

project to build a deep neural net classifier for recognizing ingredients in pictures of food

Primary LanguageJupyter Notebook

Deep ingredient recognition

status: in early development

contributors: Andrea Massari

Introduction

Dear Reader,

This is a project on ingredient recognition using deep learning by Andrea Massari. Please feel free to use any or all of its content at the condition of crediting my work whenever appropriate.

Thank you, Andrea

Credits

The idea came in collaboration with Toby Sachs-Quintana at Insight Data Science, Palo Alto, CA. The implementation scheme is largely due to work by Jingjing Chen and Chong-Wah Ngo of City University of HongKong (http://vireo.cs.cityu.edu.hk/papers/jjc_mm16.pdf).

The dataset is courtesy of allrecipes.com.

My thanks go to all who gave me their support and feedback, and in particular to Carlos Faham for the encouragement.

To-do

  • Increase dataset size
  • Preprocess more images
  • Systematize training
  • Fix tuning on Jaccard index
  • Modularize
  • Implement unit tests
  • Clean code

Future work

  • Put the model online
  • Bring in other data sources