Today's Drop features three diverse resources that present opportunities to increase knowledge in different ways. However, be prepared to close on a somewhat sad note.
Programming note: It seems the boy billionaire yanked one too many cables in his latest fit of impotency, as the Substack auto-post to Twitter broke hard this week. If you’re used to waiting for that nudge (vs subscribe) there, I’ll do my best to post manually. But, I try to limit daily bird-site interactions, since I tend to just use that platform to hold authoritarian wannabes accountable (as much as a single human can).
The GURQ
“Undo” and “redo” are idioms we've come to just expect when working on virtually anything in the digital realm.
Didn't mean to draw that line? Undo.
Hate that document edit you just made? Undo.
Actually, you didn't hate that document edit you made (a second ago) as much as you thought? Undo-undo (i.e., redo).
This functionality even followed us into the tiny glowing rectangle realm, though “shake to undo” is a pretty horrible UX.
The undo idiom goes back a ways to at least 1968 and the FRESS project at Brown University. The File Retrieval and Editing SyStem ran on an IBM 360-series mainframe running VM/CMS. It is, in retrospect, what we might call a “hypertext” system. The “undo” functionality made it easier to correct editing mistakes and greatly improved the usability of the system.
There are many ways to slay undo/redo dragons, but I found this description of “The Great Undo-Redo Quandary” to be a fun, accessible, and informative read on the topic. It both explains the quandary whilst providing a straightforward solution, and might be something you keep bookmarked (in Raindrop.io ofc) the next time one of your DIY endeavors ends up needing such functionality.
Speed Trap
I come across new “data science” discoveries/inventions on the daily. There's always some new tool, model, paper, algorithm, approach, etc. afoot, usually accompanied by some toy example. Toys are great! But, having an end-to-end, plainspoken-yet-detailed “real-life” working example to riff from is far more useful.
Sparrow Computing — a developer of computer vision tech — recently posted a cool article on how to build a vehicle speed detector with some of their FOSS libraries and PyTorch Lightning. This example ties maths with a DIY physical world setup, and does so in sufficient detail that makes it possible for virtually anyone reading this section to replicate their creation.
Here's part of their pitch:
The exciting news here is that we could make this speed detector for any traffic feed without prior knowledge about the site (no calibration required), or specialized imaging equipment (no depth sensors required). Better yet, we only needed ~300 annotated images to reach a decent speed estimate. To estimate speed, we will detect all vehicles coming toward the camera using an object detection model. We will also predict the locations of the back tire and the front tire of every vehicle using a keypoint model. Finally, we will perform object tracking to account for the same tire as it progresses frame-to-frame so that we can estimate vehicle speed.
You'll need:
A video stream of automobile traffic
~300 randomly sampled video frames from the video stream
Bounding box annotations around all the vehicles moving toward the camera
Keypoint annotations for the points where the front tire and the back tire touch the road (on the side where the vehicle is facing the camera)
The equipment required is either cheap or already in your hand (i.e., your phone), the software is open-source, and you’ll just need some patience and time, to do it up right.
Fanfare For Air Filters
While this pandemic has been all sorts of terrible, it has also been the parental unit of invention across many disciplines. I will admit to a “SQUEE!” or two when I, um, caught wind of the Corsi-Rosenthal Box that helped with both droplet spread and pollution from rampant fires (wow, the 2020s have been a real hades-scape).
The CR box wasn't the only DIY air filtration invention, and a bunch of spiffy scientists did some science on these builds. The paper is oddly accessible (even now, I find most academic papers still lean into obfuscation); and, something we all might be able to use to convince our schools and workplaces to pay more than lip service to “we care about the health/safety of all students/workers/customers/etc.”.
FIN
I have a profound affinity for old tech and every form of technology that has fostered communication. As such, I was super-bummed to read about the end of an era in Germany. ☮