Bunting Labs (YC S22)

After three days in Bratislava for the #QGIS User Conference, it's amazing to see how welcoming the QGIS community is. By far the best part of the conference was meeting so many great people from so many different industries all unified by QGIS. Everyone was excited to share their work and knowledge about QGIS—often working on things I didn't know QGIS was even capable of. Another great surprise was seeing Bunting Labs featured as the example of AI being built for QGIS—thank you to Tim Sutton for featuring us in the closing ceremony! It's clear from this conference that QGIS' importance will only grow as people invent new ways to utilize it and as it goes from being solely desktop software to being hybrid desktop and cloud software.

Just heard Brendan Ashworth speak at the QGIS User Conference about his experience building AI in QGIS! Here are the biggest takeaways: 1. The number of AI models being built for GIS will only grow 2. Hardware and software can't keep up with the pace of innovation 3. In open-source software sometimes you have to be the change you want to see The talk was recorded so we'll be posting that soon! #QGISUC2024

Hello, QGIS User Conference! Michael Egan and I are in Bratislava for the #QGISUC2024 — come say hi to us if you're here! And join us for our 11:30 talk about deploying AI models inside QGIS 😀

We've created a less expensive Personal tier at $19/month for people who only do occasional digitizing! This tier is equivalent to about 30 minutes of manual digitizing per day. If you digitize maps frequently or have heightened security requriements, our Professional and Organization tiers are likely a better fit. If you have any questions about which plan is best for you, send us an email or schedule a call on our website.

QGIS can now read and extract the metes and bounds from survey plats automatically, creating a polygon with an LLM. Typing S 21° 14' W 50x times into a coordinate geometry (CoGo) table before hitting save is so... 2023. GIS deserves better. This will be available in our QGIS plugin sometime in the next 2 days!

Introducing: the AI Georeferencer by Brendan Ashworth With the recent breakdown on Geo-referencing in my latest post using #QGIS. A much faster and cooler way has been released by Bunting Labs (YC S22) where they have developed an automatic Ai Georeferencing tool to develop GCPs (Ground Control Points) which seems to be pretty accurate, using features in your raster image and detecting similarities to data in the cloud. I am looking forward to testing this out but given the example on the link below, the accuracy seems pretty high, reducing the time taken for spatially fixing your raster data and workload as a whole. #QGIS #GIS #Raster #Georeference #WebGIS #AI #GEOAi

The first bottleneck in map digitization is georeferencing. Today, we're sharing a sneak peak at our #QGIS AI Georeferencer. To overlay a raster map into GIS software, you must add "ground control points" to connect pixels with (latitude, longitude) pairs. But what if AI could do this for you? Well... let me show you 👇 😁

I’ve been testing out the Bunting Labs (YC S22) AI Vectorizer plugin in QGIS and wanted to share my thoughts and encourage you to have a try! TLDR: If you work in the environment, geology or ecology sectors and regularly need to create vector data from raster sources check this plugin out! Whilst I'm always looking for new plugins and ways of working to introduce into my training courses, I also assist ecologists with their Habitat assessments and Biodiversity Net Gain GIS requirements. Fairly often the only information to go on is a site location PDF with a red line boundary (and sometimes not even that). Additionally, some sites are complex, like quarries that need to be digitized. I'm not always provided with a DXF to import, and they're often not un-usable without re-work. So, I’ve been taking some of my projects and seeing if the AI digitizer could help me more quickly create the data I need to speed up the process. 🌟 Installation of the plugin was easy, I got my trial key to use and using the tool is well documented and straight forwards 🌟 I used proposed site plans to create red line boundaries to send back to clients, I was very pleased how well it worked around curves and significantly sped up processing time 🌟 The new update has improved performance on non-black lines such as geology maps 🌟 When creating multiple polygons next to each other I used the snapping toolbar in conjunction with the AI digitizer to ensure they didn’t overlap. Whilst it isn’t perfect and needed some editing it was certainly faster than doing it by hand 🌟 You can correct the AI results with the shift key to assist you In conclusion, I can see this tool being especially helpful for people who are regularly digitising old maps, PDF plans or any rasters that you need vector data from. Even with any edits it’s still faster than me clicking around each of the curves and corners. The video below is where I’ve done a red line boundary and another of a geology map. This is exactly the kind of thing AI should be helping us with in the GIS community. Karolina Lehotska – you were discussing this the other day on your posts and thought you might like to know my thoughts Michael Egan – These are the current reasons for me using the digitiser and it’s been really interesting to see the tool become more accurate and useful ••••••••••••••••••••••••••••••••••••••••••• I’m Claire 👩 Founder of Maptastic 🗺. Bespoke QGIS Training from a REAL person. Delivering effective results for Environment, Ecology & Transport Sectors with detailed, effective & targeted training courses for your in-house teams. #qgis #aigis #ecologymaps #environmentmaps #geologymaps #geospatial

I've been slowly pushing out new updates to the AI Vectorizer over the last couple of weeks ... but as I pushed a bug fix for hidden raster layers today, I realized we'd crossed 30,000 downloads in #QGIS 😀

In a side by side test on the same map, our #QGIS AI Vectorizer was 2 times faster than digitizing by hand. Here’s how we ran the test and what we found: - Even compared to fast manual digitization—Brendan Ashworth averaged one vertex per second for 65 minutes straight—our AI was twice as fast, taking only 31 minutes ⏱️ - More vertices, fewer clicks: The AI created 14 times as many vertices per second while requiring 4 times fewer clicks How we ran the test: 1. Brendan found a USGS GeoTiff which the USGS has not digitized 2. He set an area on the map which he would digitize both by hand and with AI 3. He first digitized by hand in one sitting, making sure he was fully focused on getting the fastest possible manual time 4. He then digitized the same area with AI If you want to read more about this and see our evidence—including a real-time recording of the digitization, the GeoTiff, and GeoPackages of both outputs—see the blog in the comments. Brendan focused on speed over accuracy, so you may find some inaccuracies in both of the GeoPackages. This video is real-time and shows how far Brendan got by hand in the time it took to finish the same feature with AI.