Created by Carter Goldman, Alexander Glass, Donovan Chiu, Steven Bass
We all hate when we arrive at a library or a lab just to find it full, with very few study spaces open. Thus, we attempted to create a tool to show at a glance which study spaces are open.
What it does
It would have displayed a list of study spaces at Purdue, with an at-a-glance indicator of how busy each space is.
How I built it
We made an OpenCV algorithm in an AWS Lambda function which would load a live feed of a study space (in testing, live feeds of Purdue computer labs from https://www.cs.purdue.edu/resources/facilities/labs.html). We spent way too long learning OpenCV and tweaking the algorithm and did not get a working frontend this weekend.
Challenges I ran into
The algorithm did not properly identify whether or not people were in the labs. The major problem with this is that we did not have access to many feeds of study rooms, so we lacked training data for the algorithm. Thus, there were many false positives and many missed identifications. We weren't able to create any frontend because we never fully tuned the backend processing algorithms.
The URL points to my personal website; however, we believe that openstudy.space is a fantastic vanity URL for the intended project.
Accomplishments that I'm proud of
We did, however, get the algorithm working for a very small subset of the training data that we did have. We were able to look at the live feed from one computer lab in particular and determine with some accuracy where the people were.
What I learned
We learned how to use OpenCV.
What's next for Open Study Space
Actually create a frontend, tailor the model to better fit the data.