Repozytorium publikacji - Politechnika Gdańska

The article reviews neural network architectures designed for the segmentation task. It focuses mainly on instance segmentation of stacked objects. The main assumption is that segmentation is based on a color image with an additional depth layer. The paper also introduces the Stacked Bricks Dataset based on three cameras: RealSense L515, ZED2, and a synthetic one. Selected architectures: DeepLab, Mask RCNN, DEtection TRansformer, Geometry-Aware Instance Segmentation, Shapemask, Synthetic Depth Mask RCNN, Synthetic Fusion Mask RCNN (SF-Mask), Unseen Object Instance Segmentation (UOIS), Unseen Object Clustering (UOC), and You Look Only At Coefficients, have been tested on various datasets. The results show that the best architectures for stacked elements segmentation are UOIS, SF-Mask, and UOC.

Autorzy

• dr inż. Michał Czubenko link otwiera się w nowej karcie ,
• mgr inż. Artur Chrzanowski link otwiera się w nowej karcie ,
• Rafał Okuński