Bitna.AI

Inspiration

The inspirations for this project are ARTEE.AI, Rasoee, and the Billboard Next 100 Predictor.

Given these, I wanted to see if building a web application that provides a machine learning model to customize is possible. It is a fun idea to illustrate and develop, despite the challenges I came across during development.

What it does

Bitna.AI allows you to create and edit a model using PyTorch and a web interface. This is done through the Model Customization Menu under page "Model." With this menu, you can add/delete layers to the model, and Bitna will take care of adding them to the model file.

The app also lets you upload or customize datasets at your own discretion.

How we built it

The application is built with Python, Django, Vue.js, and SQLite.

Challenges we ran into

Integration of Django and Vue is quite challenging. Django has complicated topics that take a while to learn. Also, from using Vue components, it presents another challenge since .html template files are little in number and of little use.

Deploying the app on Heroku has its issues. For one, despite importing the django-heroku library, it is somehow not importing. Another is the configurations, whereupon it can get difficult to figure out what components I am missing in order to publish a live app on the site.

Accomplishments

Getting acquainted with PyTorch in a short timespan.
Learning Django basics in a short timespan.
Integrating Django and Vue.
Learning Heroku basics in a short timespan.

What we learned

Integrating PyTorch model with website has its complications, especially when using Django.
Full stack development is challenging and must take more time.