Show HN: piCards- Computer vision based audience polling solution for classrooms

52 points by AbhinavDAIICT ↗ HN
We have created a computer vision and smartphone based audience polling solution that promises all the benefits of clicker devices at minimal cost and that too without the students requiring any electronic devices.

Use of continuous feedback systems like clicker devices is proven to improve learning outcomes in classrooms and such devices are widely used in Universities. BUT such devices are costly and hence cannot be used in mid-low income setups. Also, carrying devices in bulk is a problem.

In our solution, we replace the clicker devices with normal sheets of paper with QR Code like patterns printed on them(We call the sheet — piCard). The four orientations of the sheet translate to options A, B, C and D. Each student gets one such sheet. The teacher asks a question and students rotate and lift up their sheets and show it to the teacher who has our scanning app running in his/her smartphone. The app scans all the sheets at once. Yes, you heard it right — at once! Whatever is visible in the single camera frame, gets scanned at once. The teacher gets to see the responses instantly with counts of correct and incorrect responses. These responses are synced to our analytics platform where we perform question, class, quiz, subject and student level analysis to track each individual student’s performance over time and clearly find out their strong and weak areas.

Here is a blog post related to it - https://medium.com/22bate7/picards-hello-world-3a2da6f7b885

15 comments

[ 5.1 ms ] story [ 43.1 ms ] thread
Good job. Is the code available?
Nope. We'll open source parts of it but that'll take time.
Very cool! I work in Education (Higher Ed) but I've seen K-12 use this app before: https://www.plickers.com/ which kinda offers the same functionality.

I LOVE seeing these kinds of apps come from low tech solutions combined with small bits of high tech!

Great job, and I hope you run with it.

Thanks :) Yes, Plickers also works the same way and is really awesome but for now lacks anonymous polling which I'm sure they must be working on.
Can you expand on this? I looked at the blog post and as far as I can tell piCards operate the same way as plickers. In my use of Plickers, I can either choose to give each student a specific card for personalized polling, but I can also just not identify the student with each card which allows for anonymous polling.

What have I missed here?

In plickers, you still get to see responses on the cards and since you know the students in most cases, you can identify and if you want to single out a student (which I'm sure most teachers don't do), you can do it. "piCards: Anonymous Polling App" is a truly anonymous feedback system where you cannot know the students' response unless you read their cards. This will also give more confidence to students to freely answer without worrying about getting it right.
Thanks for the awesome question :) you can get in touch with me over email as well - abhinavtripathi@22bate7.com Would love to listen to any feedback you or your fellow teachers might have. Thanks again!
These responses are synced to our analytics platform where we perform question, class, quiz, subject and student level analysis to track each individual student’s performance over time and clearly find out their strong and weak areas.

Another startup that builds dossiers on kids.

Just curious, was this based on Gwerns tweet?
I am unaware of any such tweet. Can you please direct me to that?
A tweet from 2016 or so referencing a longer writeup I made: https://www.gwern.net/Statistical-notes#proposal-hand-counti... My idea was that CNNs were good enough that you didn't have to use hacks like QR-codes or specially-designed papers; I think that someone has already done something equivalent, unless I'm thinking of you guys! (Even if you did go to CNNs for pure hand-counting, it's an obvious enough idea I doubt you got it from me, and in any case, it's implementation that counts.)