Charlie Loyd @vruba@everything.happens.horse

Adding to that repost: Do not blur (or pixelate) things to hide them. Unblurring is pretty practical in a lot of cases. If you want to hide something in a photo, the simplest, safest way is to cover it with a block of solid color.

So I think I’m borderline resolving the continental drift signal with $50 of hardware under a roof.

This makes me happy because it’s pretty comparable to the trend from a nearby science-quality GNSS receiver: https://sideshow.jpl.nasa.gov/post/links/EBMD.html

But it’s been a year, so I tested the main thing I wanted to see. If I aggressively remove outliers, average across long time periods, and eyeball a few load-bearing parameters just right – so I want to be clear that I’m not claiming rigor, only that it’s good enough for me personally – the data shows a velocity of 27.5 mm/year west, 14.4 mm/year north. (This is on the order of 1 nanometer/second.)

Its output goes straight to a postgres database on a small always-on computer. Its SSD failed a few months ago, so I lost a chunk of data, but basically it’s just recording the chip’s output. Under my roof, you might note.

One of them has been sitting on the top of the curtain rod over my desk, carefully wedged in place with magnets and a makeshift ground plane.

Last summer, I got a couple of these little USB GNSS chips, a notch above the usual consumer quality but a notch (and $200) below the RTK-capable kind: https://mou.sr/4hrR4PY

Somewhat bananas suggestion that there’s an ancient Zoroastrian necropolis in inland Madagascar: https://www.tandfonline.com/doi/full/10.1080/0067270X.2024.2380619

The magic thing about python’s sys.exit() is you don’t have to worry about importing it; if you forget, the script exits when it gets there almost exactly as if you had.

I want to map from a small input to a large output with interpolation in the output and I’m going to be able to figure it out but it’s going to feel like doing my taxes the whole time.

I have come to a place in some personal-project code where I need to use skimage.transform.warp and I keep tabbing over to the documentation, sighing, and tabbing away.

This @kissane talk is worth your time if you’re interested in pretty much anything I’m interested in. https://xoxofest.com/2024/videos/erin-kissane/

Roughly what are the odds that recognizable signs of extraterrestrial life are within our present technological and conceptual reach and we just haven’t noticed yet?

Huh. Just had a wild, cascading run of great ideas for a work project I finished like 7 years ago.

Weekend project: tinkering with cloud removal from Landsat stacks. Here the top two images are inputs and the bottom is output from them alone.

It’s strictly pixelwise and n→1 (deep set–style), so it scales to any stack depth. Notice it fills the nodata with the training set average color, and the cloud overlap with a sort of polite fog.