Comparison with state-of-the-art trackers on GOT-10k benchmark.
datasetposted on 23.11.2022, 19:00 authored by Jun Wang, Limin Zhang, Wenshuang Zhang, Yuanyun Wang, Chengzhi Deng
AO, success rates(SR)0.5 and success rates(SR)0.75 represent the average overlap and the success rate at the threshold of 0.5 and 0.75.
Read the peer-reviewed publication
multiple challenging benchmarksextensive experiments demonstrateconvolutional neural networksachieved excellent performancetracking algorithm computesnovel tracking algorithmlinear matching methodattracted extensive attentionxlink "> siameseeffectively exploit correlationsgraph attention matchingsearch region imagesgraph matchingsearch regiontracking tasktarget regionsearch regionsshuffle attentionwise informationtarget templatesrepresent targetspartial occlusionslevel informationgreatly alleviatefeature extractionart methodsappearance variations