Abstract
Dominant trackers generate a fixed-size rectangular region based on the previous prediction or initial bounding box as the model input, i.e., search region. While this manner leads to improved tracking efficiency, a fixed-size search region lacks flexibility and is likely to fail in cases, e.g., fast motion and distractor interference. Trackers tend to lose the target object due to the limited search region or be interfered by distractors due to excessive search region. In this work, we propose a novel tracking paradigm, called Search Region Regulation Tracking (SRRT), which applies a proposed search region regulator to estimate an optimal search region dynamically for every frame. To adapt the object's appearance variation during tracking, we further propose a locking-state determined updating strategy for reference frame updating. Our SRRT framework is very concise without fancy design, yet achieves evident improvements on the baselines and competitive results with other state-of-the-art trackers on seven challenging benchmarks. On the large-scale LaSOT benchmark, our SRRT improves SiamRPN++ and TransT with the absolute gains of 4.6% and 3.1% in terms of AUC.
Abstract (translated)
URL
https://arxiv.org/abs/2207.04438
