Abstract

The dataset contains the source code files used to generate a synthetic dataset with which train the data-driven model for performing the semantic segmentation of wiring harness bags. The work has been carried out in the context of the Horizon 2020 REMODEL project. Specifically, the method exploits the cut and paste technique for generating a large scale dataset of objects of interest, wiring harness bags in this case, requiring minimal human effort. The foreground images of the bags are combined with background images obtained from different sources. The method is validated performing the semantic segmentation task employing state-of-the-art deep learning models. The dataset is associated to the following publication: B. L. Žagar et al., "Copy and Paste Augmentation for Deformable Wiring Harness Bags Segmentation," 2023 IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM), Seattle, WA, USA, 2023, pp. 721-726, doi: 10.1109/AIM46323.2023.10196168.