Learning scale-variant and scale-invariant features for deep image classification

TL;DR

This paper introduces a multi-scale CNN approach that simultaneously learns scale-invariant and scale-variant features, improving image classification performance by capturing comprehensive information across multiple spatial scales.

Contribution

It presents a novel multi-scale CNN method that encourages the development of both scale-invariant and scale-variant features, addressing limitations of previous scale-invariant approaches.

Findings

Multi-scale CNN outperforms single-scale CNN on complex classification tasks.

Combined scale-invariant and scale-variant features enhance recognition accuracy.

Encouraging both feature types benefits deep image classification.

Abstract

Convolutional Neural Networks (CNNs) require large image corpora to be trained on classification tasks. The variation in image resolutions, sizes of objects and patterns depicted, and image scales, hampers CNN training and performance, because the task-relevant information varies over spatial scales. Previous work attempting to deal with such scale variations focused on encouraging scale-invariant CNN representations. However, scale-invariant representations are incomplete representations of images, because images contain scale-variant information as well. This paper addresses the combined development of scale-invariant and scale-variant representations. We propose a multi- scale CNN method to encourage the recognition of both types of features and evaluate it on a challenging image classification task involving task-relevant characteristics at multiple scales. The results show that our multi-scale CNN outperforms single-scale CNN. This leads to the conclusion that encouraging the combined development of a scale-invariant and scale-variant representation in CNNs is beneficial to image recognition performance.