A comparison of dense region detectors for image search and fine-grained classification

TL;DR

This paper compares various dense region detection methods for image search and fine-grained classification, demonstrating that alternative approaches like edge-based and super-pixel detectors outperform traditional regular dense sampling in most scenarios.

Contribution

It introduces and evaluates novel dense patch extraction techniques, including super-pixels, edges, and Zernike filters, improving state-of-the-art performance in image retrieval and classification tasks.

Findings

Edge-based patch extraction outperforms regular dense sampling.

Super-pixel and Zernike filter methods improve retrieval accuracy.

Alternative dense detectors lead to better classification results.

Abstract

We consider a pipeline for image classification or search based on coding approaches like Bag of Words or Fisher vectors. In this context, the most common approach is to extract the image patches regularly in a dense manner on several scales. This paper proposes and evaluates alternative choices to extract patches densely. Beyond simple strategies derived from regular interest region detectors, we propose approaches based on super-pixels, edges, and a bank of Zernike filters used as detectors. The different approaches are evaluated on recent image retrieval and fine-grain classification benchmarks. Our results show that the regular dense detector is outperformed by other methods in most situations, leading us to improve the state of the art in comparable setups on standard retrieval and fined-grain benchmarks. As a byproduct of our study, we show that existing methods for blob and super-pixel extraction achieve high accuracy if the patches are extracted along the edges and not around the detected regions.