Issues with Road Extraction Models from Living Atlas in ArcGIS Pro
Hello,
I'm trying to use two deep learning models from the Living Atlas for road extraction in ArcGIS Pro.
However, I encountered errors when using them with the "Detect Objects Using Deep Learning" tool.
I have attached screenshots of the errors for reference.
Model 1: Road Extraction - Global
Model 2: Road Extraction - North America
Could someone please help me understand why these errors are occurring and how I can resolve them?
Note: I am using latest ArcGIS Pro (3.3.1) and latest Deep Learning Essentials (3.3)
Accepted Solutions
Model 1: Road Extraction - Global
This issue was resolved a few months ago for this model. Please try downloading the Road Extraction - Global model again from the ArcGIS Living Atlas of the World and check if the issue persists.
Model 2: Road Extraction - North America
The Road Extraction - North America pretrained model is a pixel segmentation model, which means it is compatible with the Classify Pixels Using Deep Learning geoprocessing tool. The output of this model will be a classified raster with two classes: Road and Non-road pixels.
If you want the output roads as lines, you can use the Extract Features Using AI Models geoprocessing tool.
As shown in the screenshot above, add your raster as the Input Raster. In the Pretrained Models parameter, select Road Extraction - North America from the drop-down list.
Once you select the Road Extraction - North America model, a new section, Line Regularization Options, will appear at the bottom of the tool, as shown in the screenshot below:
You can try running the tool with the default values of Line Regularization Options and adjust the values if needed.
Model 1: Road Extraction - Global
This issue was resolved a few months ago for this model. Please try downloading the Road Extraction - Global model again from the ArcGIS Living Atlas of the World and check if the issue persists.
Model 2: Road Extraction - North America
The Road Extraction - North America pretrained model is a pixel segmentation model, which means it is compatible with the Classify Pixels Using Deep Learning geoprocessing tool. The output of this model will be a classified raster with two classes: Road and Non-road pixels.
If you want the output roads as lines, you can use the Extract Features Using AI Models geoprocessing tool.
As shown in the screenshot above, add your raster as the Input Raster. In the Pretrained Models parameter, select Road Extraction - North America from the drop-down list.
Once you select the Road Extraction - North America model, a new section, Line Regularization Options, will appear at the bottom of the tool, as shown in the screenshot below:
You can try running the tool with the default values of Line Regularization Options and adjust the values if needed.
Hi @ShivaniPathak :
Thank you for your response.
For Model 1: Road Extraction - Global, I am still facing the same issue even after downloading the latest version from the ArcGIS Living Atlas.
For Model 2: Road Extraction - North America, the solution you provided worked perfectly. The tool ran successfully, and I was able to extract the roads as lines using the Extract Features Using AI Models geoprocessing tool.
Thank you for your assistance!
Best regards,
Muhammad Elbagoury
Hi @MuhammadElbagoury ,
It's great to hear that you're now able to use the Road Extraction - North America model.
For the Road Extraction - Global model, could you try running the tool with Cell Size set to 1? You can find the cell size parameter in the Environments tab of the tool. Please refer to the screenshot shared below:
Hi @ShivaniPathak ,
Thank you again for your help.
Unfortunately, it results the same error.
But I tried something else which I'm not sure if it is right or what you want to try.
I tried to make the cell size equal to 3.28 (As my image units in Feet and this number "3.28" is equivalent to 1 meter), and it ran successfully !!!
