University of Twente Student Theses


Point cloud segmentation via active learning in the context of railway infrastructure

Hentschel, J. (2023) Point cloud segmentation via active learning in the context of railway infrastructure.

[img] PDF
Abstract:This thesis presents insights into how the active learning framework can be effectively implemented into the point cloud semantic segmentation in the railway environment. The active learning framework sets out to drastically reduce the amount of labeled data needed to train an efficient segmentation model. It achieves this by intelligently selecting new batches of data for future training through metrics that tell the algorithm how useful this data would be for the model. The overarching goal of this research is to determine how active learning compares to fully supervised learning. This will be determined by comparing their precision when running inference. In addition to this another goal is to explore the computation time required for active learning compared to fully supervised learning. The main findings in this paper conclude that active learning can provide an alternative to fully supervised learning. We were able to train models with up to 76% (86% for fully supervised) accuracy with using only 16% labeled data. Furthermore by reducing the amount of labeled data that is fed into the training process it is also possible to reduce computation times for the training process by up to 43%.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Creative Technology BSc (50447)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page