Bunting Labs (YC S22)’s Post

Transcript

Hey, Brendan here from Bunting Labs. Today I'm going to share with you the first look at our AIG reference room. You probably know us already from our AI Vectorizer, which is an autocomplete inside QGIS that lets you vectorize raster maps more easily. And today we're going to share with you a project that we've been working on for nearly six months now, and that is our AIG referencer. The AIG referencer is a QGIS plugin that will automatically generate control points for a raster that you are referencing. So I wanted to share with you a demo on two maps today. So to use the IG referencer we go to layer and open up the door referencer. With this you refer, you can load a raster normally and I have two maps and I'm going to show you the first one is a map on this is like near Alaska and Canada. It's of Mount Saint Elias and this is actually a computer rendered map. It's from a popular outdoor app called Cal Topo and basically. As opposed to manually adding control points and going back to the browser map and adding control points in the like with true latitudes and longitudes with our AIG referencer, it'll automatically add control points. And so this uploads your raster to the cloud. And using our algorithm, we actually can create control points based on features available on the map. And so this is a really dense map. There's a lot of features that are available here that it can use. For example, it's going to end up. You're referencing according to the coastline, there's lakes here, maybe perhaps the labeled peaks, for example mounting Elias, there's the border between Alaska and Canada and there's even these latitude and longitude tenmarks here. So this is a fairly straightforward not to a few reference and so we will see how it goes. This takes normally about two minutes. So I'm going to go ahead and speed it up a little bit. Great. So we just got our control points from the referencing algorithm, and you can see that these control points, because we've loaded them straight into QGIS, we can actually go in and edit them because they're a little bit wrong, but we are going to just save it and see where it is. So this was done entirely automatically and then now it's overlaid our Geotiff with the map. Now, if we evaluate accuracy, we can see, especially near the coastline, it's really good. There's not much air there. And we can even compare where the Alaska and Canada border is and you can see it's probably within like .5%. So this is pretty reasonable if we wanted a higher level of accuracy because the algorithm just generates control points. I can actually go in and move the specific control points. So that's our first example, but this has a lot of information on it. What about maps that are a little bit more difficult to do reference? Well, let's take a look at. Second map that I have. So this is actually a really old reconnaissance photo taken actually in the 1960s. And this is an airstrip in Los Angeles or outside of Los Angeles. And so this is very different from the last map. There's not meant, there's no text on the entire image. There's no hints as to its general location. And so a good question is, will the AI reference or be able to help me here? And even though this is on the other side of the spectrum, we can actually tell the IG referencer. Generally where this location is by panning on the QGIS map and so I know that this. This airstrip is in the Edwards Air Force Base outside of Los Angeles. And so actually, if I zoom in here. And then return to the AIG referencer and hit. Generate control points. It's going to use my zoomed box in QGIS to help it do your reference. And so it you can really teach, you can really use the AIG referencer as a copilot in helping you do your reference things more quickly. This is also going to take about two minutes, so let's Fast forward here. OK, so we got our control points from the dereference room. Let's take a look at what that. Ends up matching to. So let's render it. And we can go back to the QGIS to see the runway. So we can see that in the center of the control points. It has georeferenced the roads and some of the buildings here to be. Right over the target area, there's a little bit of distortion next to the runway, and I think this is because it's preferred drawing control points over here as opposed to dispersing them throughout the map. That's something that will continue to work through during our beta. The beta process will have organizations and digitizers on our pro tier. They'll be getting access to the AIG referencer first. After that, we'll be rolling out to a more general audience. So if you are interested in trying out the IG referencer, you should drop us a line. You can e-mail us. You can e-mail me at brendan@buntingla-bs.com and you can e-mail my cofounder Michael at Michael at buntingla-bs.com, and we'd be happy to hear about what you're interested in using the IG referencer for and what you thought was most cool about it. Yeah, happy digitizing.