The fastai library built on top of pytorch fascinates me and I wanted to make a tutorial detailing how to make an image classifier on Kaggle and how to deploy it. The messenger platform has a rich interface with all the features needed to deliver an app to users. It is also home to over a billion users so in essence facebook provides you the tools to build your product and the market itself.

What it does

A step by step tutorial on how to build an image classifier using pytorch/fastai and deploy it as a bot on the messenger platform

How I built it

I used the Kaggle environment to train the classifier, built a webhook with flask and hosted it on Heroku then integrated everything with the messenger platform

Challenges I ran into

I was more comfortable with fastai v1 it took a while for me to adjust the flask server to be compatible with v2 which is the latest version.

Accomplishments that I'm proud of

Completing an end to end tutorial.

What I learned

Fastai v2 and how to write a technical tutorial.

What's next for Fruit Classifier

Find other features to add to it to make it a more complete app.

Built With

Share this project: