I took some time to extend myself with some intellectual challenges. This is no holiday, but has been a lot of fun so far. Here are some highlights:
Table of contents
handWavey
If you look at nothing else; this is the one to look at.
TL;DR: A more intuitive/usable way to control your computer using a leapmotion/ultraleap controller.
There have been several implementations of this idea over the years. But I have yet to see one that I’d be happy to use on a daily basis. That’s what this project was for. I actually got it running so well that by the end of the project, I was actually faster with handWavey than I was with a mouse. To hit that home a little further, that’s comparing 4 months of usage to >30 years of usage.
I later did some informal ergonomics testing compared to a mouse, which involved not using it for a while. When I came back to it; I was a little out of practice, which really highlighted some things that needed improving. I’ve got most of the way through those, and the thing is an utter pleasure to use now.
This is my flagship project for my sabbatical, and there’s still much more to come…
Above: A shot from the video where I’m demoing handWavey, with some graphics to give it a certain movie feel.
Machine Learning Stream Cleaner
TL;DR: I developed two machine learning models. 1 to identify bad GPS/location data points. And 1 to predict what they should have been.
Above: mlsc predicting the correct value.
When I worked on my Auto Pilot project, I wrote a cool algorithm to detect bad data points. I wanted to replicate this with Machine Learning because I understood the data well. This exercise was purely academic, and for the purposes of getting a grasp of the current concepts. But it was quick to get the results that I was after, so I moved on.
The prediction was much more interesting. And there’s a lot more that I want to do with it. Especially in the the preparation of the training data set.
Cubetrix
TL;DR A 3D game a lot like 2048 where you combine matching values, but in 3 dimensions, changing the game-play in interesting ways.
In recent months, I’ve spent a lot of slack moments playing 2048, trying to get better at it (I’m not naturally good at it, so I have to work for it.) Along the lines, I started to imagine what it would be like to play in 3D where you are playing multiple games at once, where you can change your perspective. How would that behave? Now we have an answer, and it’s totally cool!
Auto-pilot
TL;DR: An auto-pilot for flying a plane, written in my own Language for robotics, Achel.
Written in my language, achel, this auto-pilot has been a long term project. Initially to show off the capabilities of achel, but has brought many interesting challenges along the way.
Resolution deep dive
TL;DR: A series of deep dives into how much resolution on your phone’s display you can actually make use of.
This was way more satisfying than I expected. I wasn’t sure if I was even going to get anything conclusive out of it, but the project just kept on giving. I hope to release more of what I found when time allows.
To help quantify various observations, I wrote an Android app.
Above: An animated series of macro photos of my test patterns on a Pentile display.
Storage deep dive
TL;DR: An old scam has just gone mainstream, and people are going to lose their data.
I’ve always considered Amazon to be one of the reasonably safer online markets, and reasonably on-top of the scams. After seeing way too many of those too-good-to-be-true listings for USB storage on Amazon, I wanted to do some of my own testing to see if the market really had advanced that quickly, or if the scammers were getting through more easily now.
This was a win-win for me:
- If the product was legitimate; I had some use-cases that would greatly benefit from cheap storage.
- If not; I’d have some interesting content to post.
Above: A listing that has been removed, and since replaced many times over.
Handheld 3D photography with a single camera
TL;DR: Long distance hand-held 3D photography with a single camera.
In the beginning, I was happy to get 1 usable tripod shot in 3 hours. By the end I was taking doing it hand-held, and nailing almost every shot within seconds.
The gist of it is to take two shots from two different horizontal positions. The trick was to focus on the right things to get a usable shot. - There are some inaccuracies that you can get away with, while there are others that you definitely can’t.
Home office
TL;DR Setting up my office to work for me.
One of the first things I did was re-do my office. It won’t be for everyone, but I love it.
I use a couple of 4k projectors mounted to an overhead bar as my main display and have everything I need close to hand, but out of the way.
There have been a lot of changes since this post because I’ve continued to evolve the design. One of the most important changes is some strategically placed shielding so that I have have the window open during the day time without getting glare on the display. I’ve also added a whole heap of sound dampening.
The result is a home office where I can attain levels of productivity that I have never gotten close to in any form of shared office space.
Much more
There’s were just some highlights. I’ve done a lot more than this.
Have a look at some of the related posts below to see more.