Research behind the platform
Maplyzer's detection and geolocalization technology comes out of applied research with the University of Vermont's AI lab and VTrans. Here's the published work behind it.
Object Tracking and Geo-localization from Street Images
Daniel Wilson, Xiaohan Zhang, Kamiku Xue, Safwan Wshah Vermont Artificial Intelligence Laboratory, Department of Computer Science, University of Vermont. VTrans contacts: Rick Scott and Alex Geller.
View Presentation SlidesEnd-to-End Sign Detection, Classification & Geo-Localization
An end-to-end system that receives road images captured from a vehicle and automatically constructs a GIS map by classifying signs and placing them at their geo-location, using an object detector, an object tracker, and an interactive visualization tool.
View Presentation SlidesGPS-RetinaNet: A Two-Stage Framework for Sign Geolocalization
A modified RetinaNet that predicts a positional offset for each sign relative to the camera, condensed into geolocalized signs with a custom tracker combining a learned metric network and a variant of the Hungarian Algorithm.
View Presentation SlidesA Large-Scale Traffic Sign Recognition Dataset
An annotated dataset of 181 distinct traffic sign classes with GPS coordinates, and an automated system to identify, classify, and geolocalize signs from roadside images.
View Presentation SlidesAutomated Traffic Sign Recognition (TSR) from Street Imagery
An automated system that classifies sign types and determines GPS location from a stream of images, alongside one of the field's early large-scale benchmark datasets for Traffic Sign Recognition.
View Presentation SlidesCase studies and technical write-ups are next
We're building out white papers and workflow write-ups alongside the research above. In the meantime, see how the VTrans partnership shaped the platform.