Accelerating Vision Transformers with Adaptive Patch Sizes

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Choudhury, Rohan, Kim, JungEun, Park, Jinhyung, Yang, Eunho, Jeni, László A., Kitani, Kris M.
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866918462482284544
author Choudhury, Rohan
Kim, JungEun
Park, Jinhyung
Yang, Eunho
Jeni, László A.
Kitani, Kris M.
author_facet Choudhury, Rohan
Kim, JungEun
Park, Jinhyung
Yang, Eunho
Jeni, László A.
Kitani, Kris M.
contents Vision Transformers (ViTs) partition input images into uniformly sized patches regardless of their content, resulting in long input sequence lengths for high-resolution images. We present Adaptive Patch Transformers (APT), which addresses this by using multiple different patch sizes within the same image. APT reduces the total number of input tokens by allocating larger patch sizes in more homogeneous areas and smaller patches in more complex ones. APT achieves a drastic speedup in ViT inference and training, increasing throughput by 40% on ViT-L and 50% on ViT-H while maintaining downstream performance, and can be applied to a previously fine-tuned ViT, converging in as little as 1 epoch. It also significantly reduces training and inference time without loss of performance in high-resolution dense visual tasks, achieving up to 30\% faster training and inference in visual QA, object detection, and semantic segmentation.
format Preprint
id arxiv_https___arxiv_org_abs_2510_18091
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Accelerating Vision Transformers with Adaptive Patch Sizes
Choudhury, Rohan
Kim, JungEun
Park, Jinhyung
Yang, Eunho
Jeni, László A.
Kitani, Kris M.
Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
Vision Transformers (ViTs) partition input images into uniformly sized patches regardless of their content, resulting in long input sequence lengths for high-resolution images. We present Adaptive Patch Transformers (APT), which addresses this by using multiple different patch sizes within the same image. APT reduces the total number of input tokens by allocating larger patch sizes in more homogeneous areas and smaller patches in more complex ones. APT achieves a drastic speedup in ViT inference and training, increasing throughput by 40% on ViT-L and 50% on ViT-H while maintaining downstream performance, and can be applied to a previously fine-tuned ViT, converging in as little as 1 epoch. It also significantly reduces training and inference time without loss of performance in high-resolution dense visual tasks, achieving up to 30\% faster training and inference in visual QA, object detection, and semantic segmentation.
title Accelerating Vision Transformers with Adaptive Patch Sizes
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2510.18091