[Will Forfang] found a app that lets you take a picture of a math equation with a phone and ask for a solution. However, the app wouldn’t read handwritten equations, so [Will] decided to see how hard that would be, using a neural network.
The results are pretty impressive (you can also see the video below). [Will] used his own handwriting on a chalkboard and had the network train on that. He also went even further and added some heuristics to identify fraction bars and infer the grouping from the relative size of the bars.
The neural net code is simple, but there are a lot of nodes. The capture image is 75×75 or 5,626 pixels in total. Each pixel has a neuron. Those feed into 50 intermediate neurons. There there is one neuron for each of the eighteen symbols the software can recognize (the digits, math operators, and so on).
The programming is just to get the neuron algorithm. The actual character recognition is handled by training the net using example input data — writing numbers over and over again on the chalkboard. Once the equation is recognized, it isn’t hard to solve the actual equation (which is not done via neural net, as far as we can tell).
We’ve seen neural nets doing speech generation, flying helicopters, and even spraying unwanted cats.
Thanks for the tip [John].
> Each pixel has a neuron
I was about to scream bloody murder how wrong that is, but then I read the article, its not even CNN. Flattened input, sigmoids, this is so 5 years old tech. Intro to state of the art today done with couple python lines:
https://www.youtube.com/watch?v=vq2nnJ4g6N0
Yeah, I almost screamed as well when they said that each “pixel has a neuron”. But it is kind of true if you are working with a VGA image, then your input layer would have 307200 neurons to match the number of inputs to map.
This is indeed very old technology as there is a single hidden layer with 20 neurons!!! hahaha. But remember that it was state of the art when Yan Lecunn made it in the 90s.
Hackaday is great for hardware, but not really much for ML, DL or AI in general.
That’s a reflection of the tip line and the time of the staff. I’m sure if you post cool projects they’ll make their way into articles.
First person to create a terminator gets a HackaDay T-shirt, and an incredible write-up.
The terminator can get his own shirt.
https://www.youtube.com/watch?v=qrMWY8Fppkk
Even for hardware the quality of the articles here are very very smooth.
Awesome lecture. I think you’ve got me started on a new interest/hobby. Thanks :-)
It may be based on older ideas but it’s still a nice little hack.
Port it to google glasses and my daughter can be the 4’th grade math wiz .
Point this App at a normal arithmetic problem and it gives you a solution.
Point this App at a Common-Core arithmetic problem, and it gives you NOTHING. (Or maybe it sends you to a Progressive-Marxist Indoctrination propaganda site while logging everything about you for further surveillance?)
Hey, this is my project. Neat! There is a video of it solving algebra, too, if you follow the link. It’s not exactly groundbreaking work, but it was fun to make.