PrunedCaps: A Case For Primary Capsules Discrimination

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Sharifi, Ramin, Shiri, Pouya, Baniasadi, Amirali
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866917128835170304
author Sharifi, Ramin
Shiri, Pouya
Baniasadi, Amirali
author_facet Sharifi, Ramin
Shiri, Pouya
Baniasadi, Amirali
contents Capsule Networks (CapsNets) are a generation of image classifiers with proven advantages over Convolutional Neural Networks (CNNs). Better robustness to affine transformation and overlapping image detection are some of the benefits associated with CapsNets. However, CapsNets cannot be classified as resource-efficient deep learning architecture due to the high number of Primary Capsules (PCs). In addition, CapsNets' training and testing are slow and resource hungry. This paper investigates the possibility of Primary Capsules pruning in CapsNets on MNIST handwritten digits, Fashion-MNIST, CIFAR-10, and SVHN datasets. We show that a pruned version of CapsNet performs up to 9.90 times faster than the conventional architecture by removing 95 percent of Capsules without a loss of accuracy. Also, our pruned architecture saves on more than 95.36 percent of floating-point operations in the dynamic routing stage of the architecture. Moreover, we provide insight into why some datasets benefit significantly from pruning while others fall behind.
format Preprint
id arxiv_https___arxiv_org_abs_2512_06003
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle PrunedCaps: A Case For Primary Capsules Discrimination
Sharifi, Ramin
Shiri, Pouya
Baniasadi, Amirali
Computer Vision and Pattern Recognition
Capsule Networks (CapsNets) are a generation of image classifiers with proven advantages over Convolutional Neural Networks (CNNs). Better robustness to affine transformation and overlapping image detection are some of the benefits associated with CapsNets. However, CapsNets cannot be classified as resource-efficient deep learning architecture due to the high number of Primary Capsules (PCs). In addition, CapsNets' training and testing are slow and resource hungry. This paper investigates the possibility of Primary Capsules pruning in CapsNets on MNIST handwritten digits, Fashion-MNIST, CIFAR-10, and SVHN datasets. We show that a pruned version of CapsNet performs up to 9.90 times faster than the conventional architecture by removing 95 percent of Capsules without a loss of accuracy. Also, our pruned architecture saves on more than 95.36 percent of floating-point operations in the dynamic routing stage of the architecture. Moreover, we provide insight into why some datasets benefit significantly from pruning while others fall behind.
title PrunedCaps: A Case For Primary Capsules Discrimination
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2512.06003