Towards Size-invariant Salient Object Detection: A Generic Evaluation and Optimization Approach

2025-09-19

AI Summary

This paper identifies a critical size-sensitivity bias in mainstream evaluation metrics for salient object detection (SOD), which systematically underestimates small yet semantically important objects in multi-scale scenarios—leading to evaluation distortion and degraded practical performance. To address this, we first provide the first theoretical analysis revealing the root cause of metric size sensitivity. We then propose SIEva, a size-invariant general framework for evaluation and optimization: it introduces a decoupled evaluation mechanism via decomposable error analysis to ensure fair measurement across scales; and it incorporates SIOpt, a model-agnostic optimization strategy compatible with diverse SOD backbones. Experiments demonstrate that SIEva significantly improves small-object detection accuracy while preserving or even surpassing original overall performance across multiple benchmarks. The implementation is publicly available.

Technology Category

• Search and Optimization: Evaluation and Analysis
• Machine Learning: Evaluation and Analysis
• Computer Vision: Bias, Fairness & Privacy

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Security and Privacy: Large-scale security measurements
• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization

Abstract

This paper investigates a fundamental yet underexplored issue in Salient Object Detection (SOD): the size-invariant property for evaluation protocols, particularly in scenarios when multiple salient objects of significantly different sizes appear within a single image. We first present a novel perspective to expose the inherent size sensitivity of existing widely used SOD metrics. Through careful theoretical derivations, we show that the evaluation outcome of an image under current SOD metrics can be essentially decomposed into a sum of several separable terms, with the contribution of each term being directly proportional to its corresponding region size. Consequently, the prediction errors would be dominated by the larger regions, while smaller yet potentially more semantically important objects are often overlooked, leading to biased performance assessments and practical degradation. To address this challenge, a generic Size-Invariant Evaluation (SIEva) framework is proposed. The core idea is to evaluate each separable component individually and then aggregate the results, thereby effectively mitigating the impact of size imbalance across objects. Building upon this, we further develop a dedicated optimization framework (SIOpt), which adheres to the size-invariant principle and significantly enhances the detection of salient objects across a broad range of sizes. Notably, SIOpt is model-agnostic and can be seamlessly integrated with a wide range of SOD backbones. Theoretically, we also present generalization analysis of SOD methods and provide evidence supporting the validity of our new evaluation protocols. Finally, comprehensive experiments speak to the efficacy of our proposed approach. The code is available at https://github.com/Ferry-Li/SI-SOD.

Problem

• Addressing size sensitivity in Salient Object Detection evaluation metrics
• Proposing size-invariant evaluation to prevent larger objects dominating assessments
• Developing optimization framework for improved detection across object sizes

Innovation

• Size-Invariant Evaluation framework SIEva
• Model-agnostic optimization framework SIOpt
• Decomposes evaluation into separable components