Using Cross-Domain Detection Loss to Infer Multi-Scale Information for Improved Tiny Head Tracking

TL;DR

This paper introduces a novel framework that combines cross-domain detection loss, multi-scale features, and small receptive fields to improve tiny head detection and tracking efficiency in crowded scenes, reducing computational costs.

Contribution

The paper presents a new integrated approach that enhances tiny head detection by bridging large and small detectors and capturing multi-scale details during training.

Findings

Improved MOTA and mAP on CroHD and CrowdHuman datasets.

Enhanced detection accuracy for tiny heads in crowded scenes.

Reduced computational requirements for head tracking.

Abstract

Head detection and tracking are essential for downstream tasks, but current methods often require large computational budgets, which increase latencies and ties up resources (e.g., processors, memory, and bandwidth). To address this, we propose a framework to enhance tiny head detection and tracking by optimizing the balance between performance and efficiency. Our framework integrates (1) a cross-domain detection loss, (2) a multi-scale module, and (3) a small receptive field detection mechanism. These innovations enhance detection by bridging the gap between large and small detectors, capturing high-frequency details at multiple scales during training, and using filters with small receptive fields to detect tiny heads. Evaluations on the CroHD and CrowdHuman datasets show improved Multiple Object Tracking Accuracy (MOTA) and mean Average Precision (mAP), demonstrating the effectiveness of our approach in crowded scenes.