SPFormer: Enhancing Vision Transformer with Superpixel Representation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Mei, Jieru, Chen, Liang-Chieh, Yuille, Alan, Xie, Cihang
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910287970435072
author Mei, Jieru
Chen, Liang-Chieh
Yuille, Alan
Xie, Cihang
author_facet Mei, Jieru
Chen, Liang-Chieh
Yuille, Alan
Xie, Cihang
contents In this work, we introduce SPFormer, a novel Vision Transformer enhanced by superpixel representation. Addressing the limitations of traditional Vision Transformers' fixed-size, non-adaptive patch partitioning, SPFormer employs superpixels that adapt to the image's content. This approach divides the image into irregular, semantically coherent regions, effectively capturing intricate details and applicable at both initial and intermediate feature levels. SPFormer, trainable end-to-end, exhibits superior performance across various benchmarks. Notably, it exhibits significant improvements on the challenging ImageNet benchmark, achieving a 1.4% increase over DeiT-T and 1.1% over DeiT-S respectively. A standout feature of SPFormer is its inherent explainability. The superpixel structure offers a window into the model's internal processes, providing valuable insights that enhance the model's interpretability. This level of clarity significantly improves SPFormer's robustness, particularly in challenging scenarios such as image rotations and occlusions, demonstrating its adaptability and resilience.
format Preprint
id arxiv_https___arxiv_org_abs_2401_02931
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle SPFormer: Enhancing Vision Transformer with Superpixel Representation
Mei, Jieru
Chen, Liang-Chieh
Yuille, Alan
Xie, Cihang
Computer Vision and Pattern Recognition
In this work, we introduce SPFormer, a novel Vision Transformer enhanced by superpixel representation. Addressing the limitations of traditional Vision Transformers' fixed-size, non-adaptive patch partitioning, SPFormer employs superpixels that adapt to the image's content. This approach divides the image into irregular, semantically coherent regions, effectively capturing intricate details and applicable at both initial and intermediate feature levels. SPFormer, trainable end-to-end, exhibits superior performance across various benchmarks. Notably, it exhibits significant improvements on the challenging ImageNet benchmark, achieving a 1.4% increase over DeiT-T and 1.1% over DeiT-S respectively. A standout feature of SPFormer is its inherent explainability. The superpixel structure offers a window into the model's internal processes, providing valuable insights that enhance the model's interpretability. This level of clarity significantly improves SPFormer's robustness, particularly in challenging scenarios such as image rotations and occlusions, demonstrating its adaptability and resilience.
title SPFormer: Enhancing Vision Transformer with Superpixel Representation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2401.02931