STEERER: Resolving Scale Variations for Counting and Localization via Selective Inheritance Learning

TL;DR

STEERER introduces a novel scale-aware object counting method that selectively inherits discriminative features across scales, significantly improving generalization and accuracy in counting and localization tasks.

Contribution

The paper proposes STEERER, a new approach with feature selection and inheritance mechanisms to better handle scale variations in object counting.

Findings

Achieves superior scale generalization on nine datasets.

Effectively improves density map quality across scales.

Outperforms existing scale-aware algorithms in counting accuracy.

Abstract

Scale variation is a deep-rooted problem in object counting, which has not been effectively addressed by existing scale-aware algorithms. An important factor is that they typically involve cooperative learning across multi-resolutions, which could be suboptimal for learning the most discriminative features from each scale. In this paper, we propose a novel method termed STEERER (\textbf{S}elec\textbf{T}iv\textbf{E} inh\textbf{ER}itance l\textbf{E}a\textbf{R}ning) that addresses the issue of scale variations in object counting. STEERER selects the most suitable scale for patch objects to boost feature extraction and only inherits discriminative features from lower to higher resolution progressively. The main insights of STEERER are a dedicated Feature Selection and Inheritance Adaptor (FSIA), which selectively forwards scale-customized features at each scale, and a Masked Selection and Inheritance Loss (MSIL) that helps to achieve high-quality density maps across all scales. Our experimental results on nine datasets with counting and localization tasks demonstrate the unprecedented scale generalization ability of STEERER. Code is available at \url{https://github.com/taohan10200/STEERER}.