CEM-FBGTinyDet: Context-Enhanced Foreground Balance with Gradient Tuning for tiny Objects

TL;DR

This paper introduces a novel architecture for tiny object detection that enhances multi-scale features, employs dynamic gradient balancing, and improves semantic representation of high-level features, leading to superior detection performance.

Contribution

The paper proposes E-FPN-BS, integrating a Context Enhancement Module and Foreground-Background Separation Module with a Dynamic Gradient-Balanced Loss for improved tiny object detection.

Findings

Outperforms existing methods on multiple benchmarks.

Effectively enhances high-level and low-level feature integration.

Demonstrates robustness and generalization across datasets.

Abstract

Tiny object detection (TOD) reveals a fundamental flaw in feature pyramid networks: high-level features (P5-P6) frequently receive zero positive anchors under standard label assignment protocols, leaving their semantic representations untrained due to exclusion from loss computation. This creates dual deficiencies: (1) Stranded high-level features become semantic dead-ends without gradient updates, while (2) low-level features lack essential semantic context for robust classification. We propose E-FPN-BS that systematically converts wasted high-level semantics into low-level feature enhancements. To address these issues, we propose E-FPN-BS, a novel architecture integrating multi-scale feature enhancement and adaptive optimization. First, our Context Enhancement Module(CEM) employs dual-branch processing to align and compress high-level features for effective global-local fusion. Second, the Foreground-Background Separation Module (FBSM) generates spatial gating masks that dynamically amplify discriminative regions. To address gradient imbalance across object scales, we further propose a Dynamic Gradient-Balanced Loss (DCLoss) that automatically modulates loss contributions via scale-aware gradient equilibrium. Extensive experiments across multiple benchmark datasets demonstrate the outstanding performance and generalization ability of our approach.