Comparison with state-of-the-art trackers on GOT-10k benchmark.

Comparison with state-of-the-art trackers on GOT-10k benchmark.

<p>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.</p>

Saved in:

Bibliographic Details
Main Author:	Jun Wang (5906) (author)
Other Authors:	Limin Zhang (297459) (author), Wenshuang Zhang (8274180) (author), Yuanyun Wang (13549920) (author), Chengzhi Deng (14164774) (author)
Published:	2022
Subjects:	Science Policy Environmental Sciences not elsewhere classified Biological Sciences not elsewhere classified Mathematical Sciences not elsewhere classified Information Systems not elsewhere classified multiple challenging benchmarks extensive experiments demonstrate convolutional neural networks achieved excellent performance tracking algorithm computes novel tracking algorithm linear matching method attracted extensive attention xlink "> siamese effectively exploit correlations graph attention matching search region images graph matching search region tracking task target region search regions shuffle attention wise information target templates represent targets partial occlusions level information greatly alleviate feature extraction art methods appearance variations
Tags:	Add Tag No Tags, Be the first to tag this record!

Be the first to leave a comment!