Quick-CapsNet (QCN): A fast alternative to Capsule Networks

2025-10-08

AI Summary

Capsule Networks (CapsNets) suffer from slow inference speeds, hindering their deployment in real-time applications. To address this, we propose Quick-CapsNet (QCN), a computationally efficient CapsNet variant that significantly accelerates inference while preserving classification accuracy. QCN achieves this by drastically reducing the number of capsules per layer and introducing a lightweight yet robust decoder, all while retaining the core dynamic routing mechanism between capsules. Crucially, QCN optimizes computation-intensive components—such as matrix multiplication and iterative routing—without compromising structural interpretability or representational fidelity. Extensive experiments on MNIST, Fashion-MNIST, SVHN, and CIFAR-10 demonstrate that QCN attains up to a 5× speedup in inference latency over the original CapsNet, with negligible accuracy degradation (<0.3% absolute drop). This work establishes a new paradigm for practical, high-efficiency capsule network deployment, bridging the gap between theoretical interpretability and real-world applicability.

Technology Category

• Computer Vision: Learning & Optimization for CV
• Machine Learning: Learning on the Edge & Model Compression
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization

Abstract

The basic computational unit in Capsule Network (CapsNet) is a capsule (vs. neurons in Convolutional Neural Networks (CNNs)). A capsule is a set of neurons, which form a vector. CapsNet is used for supervised classification of data and has achieved state-of-the-art accuracy on MNIST digit recognition dataset, outperforming conventional CNNs in detecting overlapping digits. Moreover, CapsNet shows higher robustness towards affine transformation when compared to CNNs for MNIST datasets. One of the drawbacks of CapsNet, however, is slow training and testing. This can be a bottleneck for applications that require a fast network, especially during inference. In this work, we introduce Quick-CapsNet (QCN) as a fast alternative to CapsNet, which can be a starting point to develop CapsNet for fast real-time applications. QCN builds on producing a fewer number of capsules, which results in a faster network. QCN achieves this at the cost of marginal loss in accuracy. Inference is 5x faster on MNIST, F-MNIST, SVHN and Cifar-10 datasets. We also further enhanced QCN by employing a more powerful decoder instead of the default decoder to further improve QCN.

Problem

• Addressing slow training and testing in Capsule Networks
• Reducing computational bottleneck for real-time applications
• Maintaining accuracy while significantly accelerating inference speed

Innovation

• Reduced capsule count for faster processing
• Marginal accuracy trade-off for speed enhancement
• Enhanced decoder for improved network performance