RFAssigner: A Generic Label Assignment Strategy for Dense Object Detection

2026-01-03

arXiv.org

AI Summary

This work addresses the insufficient positive sample assignment for small objects in dense object detection, which leads to imbalanced multi-scale training under existing label assignment strategies. To mitigate this issue, the authors propose RFAssigner, a novel approach that adaptively augments positive samples from candidate locations by integrating point-prior initialization with a Gaussian Receptive Field (GRF)-based distance metric. This design enables more balanced positive sample allocation across scales without requiring additional modules or heuristic rules. Evaluated on three datasets with diverse scale distributions, RFAssigner achieves state-of-the-art performance; notably, FCOS with a ResNet-50 backbone significantly outperforms existing methods across all object scales.

Technology Category

• Computer Vision: Object Detection & Categorization
• Machine Learning: Calibration & Uncertainty Quantification
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Security and Privacy: Large-scale security measurements

Abstract

Label assignment is a critical component in training dense object detectors. State-of-the-art methods typically assign each training sample a positive and a negative weight, optimizing the assignment scheme during training. However, these strategies often assign an insufficient number of positive samples to small objects, leading to a scale imbalance during training. To address this limitation, we introduce RFAssigner, a novel assignment strategy designed to enhance the multi-scale learning capabilities of dense detectors. RFAssigner first establishes an initial set of positive samples using a point-based prior. It then leverages a Gaussian Receptive Field (GRF) distance to measure the similarity between the GRFs of unassigned candidate locations and the ground-truth objects. Based on this metric, RFAssigner adaptively selects supplementary positive samples from the unassigned pool, promoting a more balanced learning process across object scales. Comprehensive experiments on three datasets with distinct object scale distributions validate the effectiveness and generalizability of our method. Notably, a single FCOS-ResNet-50 detector equipped with RFAssigner achieves state-of-the-art performance across all object scales, consistently outperforming existing strategies without requiring auxiliary modules or heuristics.

Problem

• label assignment
• dense object detection
• scale imbalance
• positive sample
• small objects

Innovation

• Label Assignment
• Dense Object Detection
• Gaussian Receptive Field
• Multi-scale Learning
• Positive Sample Selection