Making clearance outlining way easier!
Gleisurfer is an iOS app that helps the railway workers get a no-fuss clearance outline and anomalous object detection using augmented reality and machine learning.
The clearance outlining on the railroad tracks are hard to do manually with bulky spatial calliper. Today smartphones are collecting way more data than we need! We can use state-of-the-art point cloud generation, 3D reconstruction and segmentation methods to safely create the outline of the train clearance and detect if an object is not supposed to be there!
Check out our code and the output predictions for the demo provided here:
We use Pytorch and BiSeNetv2 architecture to create the railway track segmentation from a video, or a real-time flow of frames.
We add on the capability of detecting objects using yolov5 model in the frame. If the object falls into a category that does not belong to be in the tracks, the user is notified.
We rconstruct the 3D representation of a scene using structure from motion, and create meshes from the point cloud to further improve the analysis.
We ambition an iOS app where you can automatically place the train clearance outline and follow it along the tracks. The 3D reconstruction of the first run is further available in the app for the anomaly analysis and automatic detection.