Holm, Dimitrij-Marian
[Author];
Junge, Philipp
[Author];
Rutinowski, Jérôme
[Author];
Fottner, Johannes
[Author];
Herberger, David
[Author];
Hübner, Marco
[Author]
Investigation of Deep Learning Datasets for Intralogistics
- [published Version]
You can manage bookmarks using lists, please log in to your user account for this.
Media type:
E-Article;
Text
Title:
Investigation of Deep Learning Datasets for Intralogistics
Contributor:
Holm, Dimitrij-Marian
[Author];
Junge, Philipp
[Author];
Rutinowski, Jérôme
[Author];
Fottner, Johannes
[Author];
Herberger, David
[Author];
Hübner, Marco
[Author]
Published:
Hannover : publish-Ing., 2023
Published in:Proceedings of the Conference on Production Systems and Logistics: CPSL 2023 - 2 ; https://doi.org/10.15488/15326
Footnote:
Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
Description:
Deep Learning for Computer Vision has great potential in intralogistics, for example for applications such as mobile robots or autonomous forklifts. However, the availability of labelled image datasets within this area is limited. To address this problem, we benchmarked two different datasets, LOCO (Logistics Objects in Context) and the TOMIE framework (Tracking Of Multiple Industrial Entities), to figure out, if these datasets can be combined to a single one. Therefore, we examine the usability of these datasets for Object Detection tasks using the YOLOv7 framework. For this we trained several Networks and compared them with each other. A deep analysis between these two datasets shows that they are very different and only suitable for specific tasks which are not interchangeable, despite having the same domain. Deeper Investigations are done to find the reasons for this. To close the Gap between LOCO and TOMIE, a synthetic data generation pipeline for Pallets is developed and 18 000 images are rendered. Furthermore, models are trained based on the synthetic data and compared with the models trained on real data. The synthetic data generation pipeline successfully closes the reality gap, and the performance on TOMIE is increased, but the performance on LOCO is significantly weaker. To develop a deeper understanding of this behavior we examine the underlying datasets and the reasons for the performance difference are identified.