信息跟踪基准(ITB)
收藏资源简介:
信息跟踪基准(ITB)是由鹏城实验室等机构联合开发的一个小而信息丰富的视觉跟踪基准。该数据集从1.2M帧中精选出7%的帧,涵盖了所有典型的挑战性场景,旨在促进跟踪器性能的评估。ITB通过质量评估机制,考虑挑战级别、区分强度和外观变化密度,从现有基准中选择最具信息量的序列,并收集额外序列以确保跟踪场景的多样性和平衡。数据集包含180个序列,每个场景至少有20个序列,适用于评估深度跟踪器在各种实际应用场景中的表现,如智能监控、人机交互和视频分析。
The Information Tracking Benchmark (ITB) is a compact yet information-rich visual tracking benchmark jointly developed by Peng Cheng Laboratory and other institutions. This dataset selects 7% of the frames from 1.2 million frames, covering all typical challenging scenarios, and aims to facilitate the evaluation of tracker performance. ITB selects the most informative sequences from existing benchmarks through a quality assessment mechanism that takes into account challenge level, discrimination intensity and appearance variation density, and collects additional sequences to ensure the diversity and balance of tracking scenarios. The dataset contains 180 sequences, with at least 20 sequences per scenario, and is suitable for evaluating the performance of deep trackers in various practical application scenarios such as intelligent surveillance, human-computer interaction and video analysis.

- 1An Informative Tracking Benchmark鹏城实验室 · 2021年



