LaSOT: A High-Quality Benchmark for Large-Scale Single Object Tracking

Abstract

In this paper, we present LaSOT, a high-quality benchmark for Large-scale Single Object Tracking. LaSOT consists of 1,400 sequences with more than 3.5M frames in total. Each frame in these sequences is carefully and manually annotated with a bounding box, making LaSOT the largest, to the best of our knowledge, densely annotated tracking benchmark. The average video length of LaSOT is more than 2,500 frames, and each sequence comprises various challenges deriving from the wild where target objects may disappear and re-appear again in the view. By releasing LaSOT, we expect to provide the community with a large-scale dedicated benchmark with high quality for both the training of deep trackers and the veritable evaluation of tracking algorithms. Moreover, considering the close connections of visual appearance and natural language, we enrich LaSOT by providing additional language specification, aiming at encouraging the exploration of natural linguistic feature for tracking. A thorough experimental evaluation of 35 tracking algorithms on LaSOT is presented with detailed analysis, and the results demonstrate that there is still a big room for improvements.

Visual Tracking by Means of Deep Reinforcement Learning and an Expert Demonstrator

“…The proposed network was then trained inside an RL on-policy asynchronous Actor-Critic framework [32] that incorporated parallel SL agents. Experiments showed that the proposed A3CT and A3CTD trackers outperform state-of-theart methods on the most recent the GOT-10k test set [18], LaSOT [9], UAV123 [35] benchmarks, and perform comparably with the state-of-the-art on OTB-2015 [50] and VOT benchmarks [26]. Moreover, A3CT and A3CTD achieved a processing speed of 90 and 50 FPS respectively and thus are suitable for real-time applications.…”

“…In this section we report the experimental setup and we discuss the results, obtained by the proposed trackers A3CT and A3CTD, on the benchmarks GOT-10k [18], LaSOT [9], UAV123 [35], OTB-100 [50], VOT-2018 [26] and VOT-2019.…”

“…Visual Tracking has received increasing interest thanks to the introduction of new benchmarks [50,9,35,18] and challenges [28,23,24,25,26].…”

“…We performed evaluations of A3CT and A3CTD performance on the test set of LaSOT benchmark [9]. This dataset is composed of 280 videos with a total of more than 650k frames and an average sequence length of 2500 frames.…”

“…A state-of-theart tracking algorithm [2] is run on a large-scale tracking dataset [18] to obtain the demonstrations. Experiments will show that the proposed A3CT and A3CTD trackers perform comparably with state-of-the-art methods on the GOT-10k test set [18], LaSOT [9], UAV123 [35], OTB-100 [50] and VOT benchmarks [24,26], while achieving a processing speed of 90 FPS and 50 FPS respectively.…”