Contribution

Towards Automated Handling and Sorting of Garments combining Visual Language Models and Convolutional Neural Networks

---

Abstract

Ambitious goals set by the European Union are aiming towards full recycle-ability of garments by 2030. According to the EU, 12 kg of garments are discarded by each citizen per year. In order to process such vast amounts of garments,automation of garment handling and recycling is unavoidable. Automated handling and sorting of such garments is a Major challenge in the field of robotics. Current approaches specialize in one part of this challenging task. For sorting, current approaches use cameras and pre-trained networks with a dataset with a pre-defined set of classes. This paper presents an approach of using artificial intelligence (Convolutional Neural Network and Visual Language Models) to locate and separate garments from a pile and identifying and sorting them into dedicated containers. This combines the advantages of both neural network types, where convolutional neural networks are used for grasping (segmentation and corner detection) and visual language models are used for classification of garment types and to help the grasp prediction network in narrowing in on better grasp positions.

Keywords: Garments, Sorting, Visual Language Models

How to Cite:

Ergun, S., Mitterer, T. & Zangl, H., (2025) “Towards Automated Handling and Sorting of Garments combining Visual Language Models and Convolutional Neural Networks”, ARW Proceedings 25(1), 25-30. doi: https://doi.org/10.34749/3061-0710.2025.4