%0 Journal Article
%T Multi-scale hybrid vision transformer and Sinkhorn tokenizer for sewer defect classification
%A Joakim Bruslund Haurum
%A Meysam Madadi
%A Sergio Escalera
%A Thomas B. Moeslund
%J Automation in Construction
%D 2022
%V 144
%F Joakim Bruslund Haurum2022
%O HuPBA
%O exported from refbase (http://refbase.cvc.uab.es/show.php?record=3780), last updated on Tue, 25 Apr 2023 15:26:25 +0200
%X A crucial part of image classification consists of capturing non-local spatial semantics of image content. This paper describes the multi-scale hybrid vision transformer (MSHViT), an extension of the classical convolutional neural network (CNN) backbone, for multi-label sewer defect classification. To better model spatial semantics in the images, features are aggregated at different scales non-locally through the use of a lightweight vision transformer, and a smaller set of tokens was produced through a novel Sinkhorn clustering-based tokenizer using distinct cluster centers. The proposed MSHViT and Sinkhorn tokenizer were evaluated on the Sewer-ML multi-label sewer defect classification dataset, showing consistent performance improvements of up to 2.53 percentage points.
%K Sewer Defect Classification
%K Vision Transformers
%K Sinkhorn-Knopp
%K Convolutional Neural Networks
%K Closed-Circuit Television
%K Sewer Inspection
%U http://dx.doi.org/10.1016/j.autcon.2022.104614
%P 104614