A Comparison of CNN and Classic Features for Image Retrieval

TL;DR

This paper compares CNN-based and traditional features for image retrieval, analyzing their effectiveness in different contexts through experiments on keypoint detection, descriptors, and a bag-of-words retrieval model.

Contribution

It provides a comprehensive comparison of CNN and classic features for image retrieval, highlighting their strengths in various scenarios.

Findings

CNN-based methods excel in certain contexts

Traditional features perform better in others

Performance varies depending on the application scenario

Abstract

Feature detectors and descriptors have been successfully used for various computer vision tasks, such as video object tracking and content-based image retrieval. Many methods use image gradients in different stages of the detection-description pipeline to describe local image structures. Recently, some, or all, of these stages have been replaced by convolutional neural networks (CNNs), in order to increase their performance. A detector is defined as a selection problem, which makes it more challenging to implement as a CNN. They are therefore generally defined as regressors, converting input images to score maps and keypoints can be selected with non-maximum suppression. This paper discusses and compares several recent methods that use CNNs for keypoint detection. Experiments are performed both on the CNN based approaches, as well as a selection of conventional methods. In addition to qualitative measures defined on keypoints and descriptors, the bag-of-words (BoW) model is used to implement an image retrieval application, in order to determine how the methods perform in practice. The results show that each type of features are best in different contexts.