Zhenlong Li’s Post

I'm excited to share our recent research on autonomous GIS with the development of a "GIS Copilot" that allows users to perform spatial analysis in QGIS using natural language for both vector and raster data. The Copilot was tested with more than 100 spatial analysis tasks spanning various complexity levels. The result shows that the GIS Copilot has strong potential in automating foundational GIS operations, with a high success rate for basic and intermediate tasks, while challenges remain in achieving full autonomy for more complex tasks without explicit guidance. This work represents a promising step toward the development of autonomous GIS, aiming to make spatial analysis more accessible to all. To learn more, please check out our preprint article at https://lnkd.in/euPQphf3 The GIS Copilot can be downloaded from the official QGIS plugin page at https://lnkd.in/eGjBzS-w Video demonstrations: https://lnkd.in/ejt7tZ94 The 110 test cases/analysis examples can be found at https://lnkd.in/eKUXWA8h https://lnkd.in/eAKu8xR8 The system is open sourced at https://lnkd.in/eFrdmjBk Temitope Ezekiel Akinboyewa, Zhenlong Li, Huan Ning, M. Naser Lessani #artificialintelligence #gis #giscience #geoai #spatialanalysis #QGIS #GDAL #autonomous #largelanguagemodels #chatgpt #generativeAI #agent

I'm excited to share our recent research on autonomous GIS with the development of a "GIS Copilot" that allows users to perform spatial analysis in QGIS using natural language for both vector and raster data. The Copilot was tested with more than 100 spatial analysis tasks spanning various complexity levels. The result shows that the GIS Copilot has strong potential in automating foundational GIS operations, with a high success rate for basic and intermediate tasks, while challenges remain in achieving full autonomy for more complex tasks without explicit guidance. This work represents a promising step toward the development of autonomous GIS, aiming to make spatial analysis more accessible to all. To learn more, please check out our preprint article at https://lnkd.in/euPQphf3 The GIS Copilot can be downloaded from the official QGIS plugin page at https://lnkd.in/eGjBzS-w Video demonstrations: https://lnkd.in/ejt7tZ94 The 110 test cases/analysis examples can be found at https://lnkd.in/eKUXWA8h https://lnkd.in/eAKu8xR8 The system is open sourced at https://lnkd.in/eFrdmjBk Temitope Ezekiel Akinboyewa, Zhenlong Li, Huan Ning, M. Naser Lessani #artificialintelligence #gis #giscience #geoai #spatialanalysis #QGIS #GDAL #autonomous #largelanguagemodels #chatgpt #generativeAI #agent

Our #QGIS AI Georeferencer is now available! You can now georeference north oriented aerial imagery and maps automatically in QGIS—here's how 👇 First, go to the Manage and Install Plugins window in QGIS and either update your Bunting Labs AI Vectorizer plugin to the latest version or install it if you haven't already Then, go to Layer -> Georeferencer and you will find two new buttons 🛰️ Aerial imagery georeferencing—our most powerful georeferencer. To use this: 1. load your satellite or aerial image in the georeferencer, 2. zoom your QGIS window to be roughly the same area as the image, 3. select the satellite icon "Generate GCP points from satellite" to generate control points 🌐⚡️ Full georeferencing—our experimental georeferencer. This mode uses our AI model built to read the map, identify features that a human would georeference to like roads, building outlines, and coordinates, and place ground control points. This allows our AI to georeference maps that do not have an aerial imagery base. To use this: 1. load your map in the georefernecer, 2. zoom your QGIS window to be roughly the same area as the map, 3. select the globe and lightning bolt icon "Generate GCP points with AI" to generate control points Make sure that whatever you try to georeference is north oriented. But, we are training it to handle rotated maps so that will be coming soon!

Our #QGIS AI Georeferencer is now available! You can now georeference north oriented aerial imagery and maps automatically in QGIS—here's how 👇 First, go to the Manage and Install Plugins window in QGIS and either update your Bunting Labs AI Vectorizer plugin to the latest version or install it if you haven't already Then, go to Layer -> Georeferencer and you will find two new buttons 🛰️ Aerial imagery georeferencing—our most powerful georeferencer. To use this: 1. load your satellite or aerial image in the georeferencer, 2. zoom your QGIS window to be roughly the same area as the image, 3. select the satellite icon "Generate GCP points from satellite" to generate control points 🌐⚡️ Full georeferencing—our experimental georeferencer. This mode uses our AI model built to read the map, identify features that a human would georeference to like roads, building outlines, and coordinates, and place ground control points. This allows our AI to georeference maps that do not have an aerial imagery base. To use this: 1. load your map in the georefernecer, 2. zoom your QGIS window to be roughly the same area as the map, 3. select the globe and lightning bolt icon "Generate GCP points with AI" to generate control points Make sure that whatever you try to georeference is north oriented. But, we are training it to handle rotated maps so that will be coming soon!

Our #QGIS AI Georeferencer is now available! You can now georeference north oriented aerial imagery and maps automatically in QGIS—here's how 👇 First, go to the Manage and Install Plugins window in QGIS and either update your Bunting Labs AI Vectorizer plugin to the latest version or install it if you haven't already Then, go to Layer -> Georeferencer and you will find two new buttons 🛰️ Aerial imagery georeferencing—our most powerful georeferencer. To use this: 1. load your satellite or aerial image in the georeferencer, 2. zoom your QGIS window to be roughly the same area as the image, 3. select the satellite icon "Generate GCP points from satellite" to generate control points 🌐⚡️ Full georeferencing—our experimental georeferencer. This mode uses our AI model built to read the map, identify features that a human would georeference to like roads, building outlines, and coordinates, and place ground control points. This allows our AI to georeference maps that do not have an aerial imagery base. To use this: 1. load your map in the georefernecer, 2. zoom your QGIS window to be roughly the same area as the map, 3. select the globe and lightning bolt icon "Generate GCP points with AI" to generate control points Make sure that whatever you try to georeference is north oriented. But, we are training it to handle rotated maps so that will be coming soon!

Our #QGIS AI Georeferencer is now available! You can now georeference north oriented aerial imagery and maps automatically in QGIS—here's how 👇 First, go to the Manage and Install Plugins window in QGIS and either update your Bunting Labs AI Vectorizer plugin to the latest version or install it if you haven't already Then, go to Layer -> Georeferencer and you will find two new buttons 🛰️ Aerial imagery georeferencing—our most powerful georeferencer. To use this: 1. load your satellite or aerial image in the georeferencer, 2. zoom your QGIS window to be roughly the same area as the image, 3. select the satellite icon "Generate GCP points from satellite" to generate control points 🌐⚡️ Full georeferencing—our experimental georeferencer. This mode uses our AI model built to read the map, identify features that a human would georeference to like roads, building outlines, and coordinates, and place ground control points. This allows our AI to georeference maps that do not have an aerial imagery base. To use this: 1. load your map in the georefernecer, 2. zoom your QGIS window to be roughly the same area as the map, 3. select the globe and lightning bolt icon "Generate GCP points with AI" to generate control points Make sure that whatever you try to georeference is north oriented. But, we are training it to handle rotated maps so that will be coming soon!