sagarr / SeeFood

Shazam for Food
1 stars 2 forks source link
convnet food-classification machine-learning silicon-valley

See Food

Shazam for Food, an idea shamelessly copied from Silicon Valley S4Ep3

data creation

The biggest quest in such application is data creation. I have used Fatun Chrome Extension which downloads images from google image search page for various food items.

training

Place food images in /data dir with food name folder as name and train with below command $ python src/train.py

the weights get stored in 'weights' dir.

prediction

Run

$ cd src
$ python3 predict.py ../data/val/vadapav/images\ \(24\).jpg

and it will print food class name

Webapp

Run flask webapp

$ cd src
$ export FLASK_APP=webapp/webapp.py
$ python3 -m flask run