SM+: Refined Scale Match for Tiny Person Detection

TL;DR

This paper introduces SM+ a refined scale matching method for tiny person detection that improves dataset scale alignment and enhances detection performance in large images, addressing a key challenge in small object detection.

Contribution

The paper proposes SM+ for better scale alignment at the instance level and introduces PSI for background preservation, advancing tiny object detection techniques.

Findings

SM+ significantly improves detection accuracy on TinyPerson dataset.

SM+ outperforms state-of-the-art detectors with notable margins.

The method effectively enhances scale similarity between pre-training and target datasets.

Abstract

Detecting tiny objects ( e.g., less than 20 x 20 pixels) in large-scale images is an important yet open problem. Modern CNN-based detectors are challenged by the scale mismatch between the dataset for network pre-training and the target dataset for detector training. In this paper, we investigate the scale alignment between pre-training and target datasets, and propose a new refined Scale Match method (termed SM+) for tiny person detection. SM+ improves the scale match from image level to instance level, and effectively promotes the similarity between pre-training and target dataset. Moreover, considering SM+ possibly destroys the image structure, a new probabilistic structure inpainting (PSI) method is proposed for the background processing. Experiments conducted across various detectors show that SM+ noticeably improves the performance on TinyPerson, and outperforms the state-of-the-art detectors with a significant margin.