RGBT234-Miss, LasHeR245-Miss, VTUAV176-Miss

本研究构建了三个高质量的RGBT跟踪基准数据集：RGBT234-Miss、LasHeR245-Miss和VTUAV176-Miss，用于评估在多种模态缺失场景下的跟踪性能。这些数据集通过模拟真实世界中的挑战，考虑了不同的模态缺失模式和缺失比例，如长时间缺失、切换缺失和随机缺失等。数据集的构建基于现有的模态完整数据集，通过设计层次化的组合方案来模拟可能的模态缺失场景。这些数据集的建立为研究模态缺失下的RGBT跟踪问题提供了全面的评估平台，旨在解决模态信息缺失对跟踪性能的影响问题。

This study constructs three high-quality RGBT tracking benchmark datasets: RGBT234-Miss, LasHeR245-Miss, and VTUAV176-Miss, which are dedicated to evaluating tracking performance under various modality missing scenarios. These datasets simulate real-world tracking challenges and incorporate diverse modality missing patterns and missing ratios, including long-term missing, modality switching missing, and random missing. Developed based on existing fully-modal complete datasets, the proposed datasets adopt hierarchical combination schemes to simulate plausible modality missing scenarios. The establishment of these datasets provides a comprehensive evaluation platform for researching RGBT tracking problems under modality missing conditions, aiming to address the adverse impacts of missing modal information on tracking performance.

搜集汇总

背景与挑战

背景概述

该数据集包含RGBT234-Miss、LasHeR245-Miss和VTUAV176-Miss三个基准数据集，专门用于RGBT（可见光-热红外）跟踪中的模态缺失研究。这些数据集模拟了真实场景中因传感器校准或数据传输错误导致的模态信息缺失问题，旨在评估跟踪模型在部分模态缺失下的鲁棒性，并支持可逆提示学习等方法的开发。