Create LRS Events From Street Level Imagery

Using the deep learning capabilities in ArcGIS Pro and the ArcGIS Python API, we developed a tool that will automatically create LRS event features given street level imagery of street signs.

Comment

Inspiration

We were inspired to work on this project to explore the potential value that customers could have by fully integrating an LRS with deep learning. We hoped that this tool would assist users by automating the often arduous event editing process.

What it does

The tool can be run in the Geoprocessing pane in ArcGIS Pro. It accepts point features with a raster field containing street level imagery. The tool then runs these images through a custom trained deep learning model that classifies street signs (Stop signs, Speed Limit signs, and Other signs.) The tool fully populates all required fields according to the LRS event schema.

How we built it

This tool was built using ArcGIS Pro 2.9 and the Python API.

Challenges we ran into

One of the biggest challenges we ran into was a lack of training data. The model we developed often inaccurately classifies street signs, though this could be mitigated easily given more time to develop the deep learning model.

Accomplishments that we're proud of

We proved the concept! This could turn into a huge value add for customers and make them more successful in their work.

What we learned

We did a lot of exploration into the capabilities of deep learning in ArcGIS Pro.

What's next for Create LRS Events From Street Level Imagery

We hope to productize this as part of ArcGIS Roads and Highways! We also hope to read information off of road signs and populate that information onto event features.

Built With

Share this project:

Updates