Rain Cloud Image Classifier

A progressive web app that allows the user to take a pic of a cloud and detect whether it is going to rain or not.

Comment

Inspiration:

A very unique group of first-time hackers with extremely varying degrees of skill came together to try to do something cool for TigerHacks. We wanted to try and do something with open source machine learning library TensorFlow, and also with Progressive Web Apps. An idea came into our heads and we went hard to work. One thing to note is that none of our group had any ML or PWA experience. And when I say that half of our group did not have experience, I mean that some of their first text-editors were installed this weekend.

Goal:

To be able to take a picture of a cloud on your phone or computer, run it against a TensorFlow image classifier using a convolutional neural network, and have that classifier tell you if it thinks it is a rain cloud or not.

Workflow:

With none of us having any ML experience, and most of us having next to no web dev or real life experience, we split our group into two sub-teams. Those who wanted to work on the ML back-end and those wanted to work on the progressive web-app front end. After getting the ML image classifier retrained using hundreds of images and the progressive web-app front-end set up, we started to seek out how we would connect everything. We decided that a Flask API would be the way to go.

Result

All of our individual parts do what we want them to do. The front end looks great and does what we want it to do, and still follows the principles of a progressive web app. Our image classifier can detect rain clouds with over a 99% success rate. Lastly our Flask API is working...but we cannot seem to connect all of our pieces. We are still extremely proud of what we have accomplished this weekend, especially considering only three of our nine members came into this team with a Github account, and five of the nine installed their first text editors on Friday.

We would also like to give a shout-out to Sam Kreter for sparing a large amount of his time helping us with Docker.

Share this project:

Updates