This era of Zoom classes has been tough for everyone. We have noticed that even the most beloved, experienced, and on-top-of-it professors can very easily get overwhelmed trying to not only ensure their technology works but also make their students are content, learning, and keeping up.

So, we created, which analyzes students' emotions - whether that be stress, surprise, happiness, or more. Our platform tells teachers the overall sentiments of their class - were students confused? Exhausted? Inspired? - so they can have a better idea of how their class is learning.

In this age of learning how to adapt to our remote environment, we hope gives instructors a less stressful, productive, and efficient way to evaluate how their class is learning.

What it does is a web app that uses machine learning to analyze the sentiments of people's faces in real-time. At this moment, we focused on stress analytics but there is opportunity for much more. then creates a digestible report of its sentiment analysis, so instructors can learn from the data.

How we built it

We primarily spent our time on the backend. We used Google Colab as our IDE, and we used MIT's open-source Facial Expression Recognition model (FER): link. We used some other essential Python libraries and frameworks, like OpenCV to capture video in real-time and Matplotlib to graph our data in a digestible form.

Challenges we ran into

We ran out of time to create a full-fledged frontend product. Our original idea was to create a user-friendly dashboard where users could access all the analytics of their class - from graphs, to trends, to possibly suggestions. Our backend took a little longer than we thought (which was why we didn't get around to the frontend), but that overall went pretty smoothly.

We did have a plan for our frontend though - we wanted to create a JS Web App with a very modern style and actually started the code (although we didn't get very far).

Accomplishments that we're proud of

We're really proud that we got the backend / machine learning part of it to work! There are definitely a lot of moving pieces to the backend, so we were very proud when we were able to turn on our webcam and see the analysis happen in real time!

It was also an amazing experience working together. All of us on our team got closer over the past few days, and we feel really lucky that we were not only able to spin up a working product but also become better friends - we're looking forward to doing more hackathons together in the future!

What we learned

For a couple of us, this was our first time using FER ML models, so it was a really great experience getting to successfully work with one. This was a great opportunity for many of us to practice Python as well, as our school primarily teaches in Java!

Also, although we didn't end up having time to finish up the frontend, we put a lot of thought into designing it - from storyboarding to reflecting on the most user-friendly processes, and we all realized how much thought goes into creating a truly pleasant-to-use interface and look forward to applying that thinking in future projects.

What's next for Insaightful

Our immediate next step is to create the frontend product that we designed: we are stoked that our analytics work, but now we want to figure out how to best display those analytics in a way that would benefit teachers the most!

We also want to figure out more ways to conduct analytics. We understand that, as students, we may not be aware of some other functionality that instructors may want, so our best bet would be to conduct some user research and really figure out what analytics and features should be integrated into our product.


We are competing in the Tools track.

Built With

Share this project: