Partitioned Saliency Ranking with Dense Pyramid Transformers

TL;DR

This paper introduces a novel ranking by partition paradigm and Dense Pyramid Transformer to improve saliency ranking accuracy by reducing ambiguity and enhancing feature interactions across scales.

Contribution

It proposes the ranking by partition paradigm and Dense Pyramid Transformer, advancing saliency ranking methods with improved performance and reduced ambiguity.

Findings

Outperforms existing saliency ranking methods

Reduces ranking ambiguities effectively

Enhances feature interactions with Dense Pyramid Transformer

Abstract

In recent years, saliency ranking has emerged as a challenging task focusing on assessing the degree of saliency at instance-level. Being subjective, even humans struggle to identify the precise order of all salient instances. Previous approaches undertake the saliency ranking by directly sorting the rank scores of salient instances, which have not explicitly resolved the inherent ambiguities. To overcome this limitation, we propose the ranking by partition paradigm, which segments unordered salient instances into partitions and then ranks them based on the correlations among these partitions. The ranking by partition paradigm alleviates ranking ambiguities in a general sense, as it consistently improves the performance of other saliency ranking models. Additionally, we introduce the Dense Pyramid Transformer (DPT) to enable global cross-scale interactions, which significantly enhances feature interactions with reduced computational burden. Extensive experiments demonstrate that our approach outperforms all existing methods. The code for our method is available at \url{https://github.com/ssecv/PSR}.