Michael Skaug

@pwang @bradneuberg @alexgleith But in general I think using ML to learn a transform from low to high resolution is a good idea. Especially if the imaging system and conditions are not simple.

@pwang @bradneuberg @alexgleith What worries me about this paper is that there was no ground truth data and the accuracy can’t really be evaluated

Interesting article on using the 12 Sentinel-2 bands at 10 meter resolution to super resolve them to 1 meter resolution @ya_71389/sentinel-2-deep-resolution-3-0-c71a601a2253" target="_blank" rel="nofollow noopener">medium.com/@ya_71389/sent…

@mouthofmorrison @kevinakwok Array Labs post?

@kevinakwok Occasionally punctuated by reading someone else’s deep analysis of a niche you have studied for years and not having the energy to rebut it

Life is just having an ever growing list of very niche topics you wish someone would spend months doing an analysis of. and knowing probably no one including yourself will

@aravindEO To what extent does a traditional GIS address this? Like ESRI or QGIS? Do they not have the scale you’re thinking of? Too specialized to domain experts?

One idea that I want to see come to fruition in EO: A data/sensor/provider-agnostic platform to enable the pre-processing, processing & visualisation of geospatial data to make EO accessible, usable & fusable, enabling application building. A Unity engine for EO, if you will.

@SeanGorman @fikocian Huh. That makes me wonder/worry about some “truth” labeling I’ve done in the past.

@fikocian Thanks! The challenge is if you are using an imagery provider like Maxar the default CE90 is 5m (90% of the error is less than 5m). So your GPS point could be very accurate but the map makes it look off. The question is which map tile provider has the least amount of these errors

I was curious if anyone knows which basemap service's tiles have the least amount of geodetic error. For instance if you have centimeter level RTK GNSS measurements, which map service will do the best job aligning to reality?

@MaxLenormand See you there!

I’m so stupidly excited to be going to SatCamp I can’t wait to meet so many people in this weird little amazing community of geospatial nerds, and discover a totally new part of the US

@bode_brad @aravindEO @Amphinicy Great points. Do you think open source or its absence plays a role? In some fields OSS can drive adoption of common tools. But I imagine that export control might make open source hard.

@mike_skaug @aravindEO 3) Digitized formats and interfaces are not exactly incentivized by the hardware manufacturers. They would much rather have vendor lock in than provide "as a service". Props to certain companies like @Amphinicy for their forward thinking.

Although I spend my time predominantly in Earth observation, one area of the space industry that interests me a lot (given my software past) is creating and owning the software stack for space. Immense potential for big recurring revenues and impact in the industry!

@aravindEO Yes! A lot of effort and bespoke software in places that are not the value add for a company. I think @bode_brad might be working on this

Historically software for many of these segments have been too bespoke in nature, so I am not entirely sure if it is possible for one company to own all of these areas, but I do know there are several companies focusing on most of them.

@opencholmes @HarelDan @OvertureMaps It sounds like the data is in parquet files on s3 so you could probably download them directly if you want. I think the docs explain that as an option.

@HarelDan @OvertureMaps I find it open enough for a first release, and appreciate they didn’t hold it back for longer. But I agree it shouldn’t lead with ‘sign up for Athena’ I think the core format is decent, just needs the ecosystem tools.

This could be big... Even huge? Overture Maps releases a global POI open dataset, an aggregate of Microsoft @bingmaps and @Meta places data. Read more: overturemaps.org/overture-maps-… #openstreetmap #opendata #gischat

This is useful. Although it would be easier to remember if it was just ‘aws ec2 ssh’

This is an interesting dashboard from @RockyMtnInst comparing the operating characteristics of current and planned satellites for methane emissions monitoring - spect.rmi.org

Even though I knew lossless compression was a thing, it always seemed kind of oxymoronic. This article helped me understand how it works. I can imagine why even a “random” sequence has some more or less common elements so could be compressed slightly.

They also made a replica of the Matterhorn that’s 15 nanometers high

The smallest map ever created? It might this one made at IBM research Zurich using thermal scanning probe lithography. It’s about 10 microns wide

@DesignMuseum Reminds me of the buzzers in Venice

That feeling when the Sunday scaries hit 😮 A look at architect Kazumasa Yamashita’s unique Face House in Kyoto, Japan. Built in 1974, each part of the facade is designed to mimic functions of an actual face; windows as eyes, air vents inside the nostrils. 📸 Stefano Perego

@dkwiens There seems to be some python or goal based tools for viewshed (like gdal.org/programs/gdal_…) but going from viewshed to horizon line doesn’t seem obvious. It’s kind of like the concave hull of the viewshed?

@dkwiens That’s an interesting extension of the viewshed problem. Given the viewshed, you want to find the most distant point. I’ve been working on a viewshed implementation, but it’s not python and it’s not done, so not much help to you. There must be a python version somewhere.

Oh, wow. If you get to the point of doing this, you're in big trouble. And the manufacturer's warning is probably the last thing on your mind.