LaSOT: A High-Quality Benchmark for Large-Scale Single Object Tracking
By Heng Fan, Liting Lin, Fan Yang, Peng Chu, Ge Deng, Sijia Yu, Hexin Bai, Yong Xu, Chunyuan Liao, Haibin Ling. 2018
1,585 citations recorded by the source corpus.
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.
Subfields: Computer Vision and Pattern Recognition, Safety, Risk, Reliability and Quality, Evaluation & Benchmarks
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Ming-Hsuan Yang
- Dacheng Tao
- Xiaogang Wang
- Jitendra Malik
- Trevor Darrell
- Kristen Grauman
- Andrew Zisserman
In Safety, Risk, Reliability and Quality
- Martin Danelljan
- Houwen Peng
- Zhipeng Zhang
- Goutam Bhat
- Fahad Shahbaz Khan
- Qiang Wang
- Luc Van Gool
- Haibin Ling
In Evaluation & Benchmarks
- Kristen Grauman
- Jitendra Malik
- Martin Danelljan
- Luc Van Gool
- Antonio Torralba
- Giovanni Maria Farinella
- Antonino Furnari
- Ming-Hsuan Yang
Citations in the corpus
- Ego4D: Around the World in 3,000 Hours of Egocentric Video (2021, background): “[62] Heng Fan, Haibin Ling, Liting Lin, Fan Yang, Peng Chu, Ge Deng, Sijia Yu, Hexin Bai, Yong Xu, and Chunyuan Liao.”
- Siam R-CNN: Visual Tracking by Re-Detection (2019, methodology): “Except where specified otherwise, we train Siam R-CNN on the training sets of multiple tracking datasets simultaneously: ImageNet VID [55] (4000 videos), YouTube-VOS 2018 [72] (3471 videos), GOT-10k [30] (9335 videos) and LaSOT [18] (1120 videos).”
- How to Train Your Energy-Based Model for Regression (2020, methodology): “We also evaluate on the LaSOT [11] test set, containing 280 long videos (2 500 frames on average).”
- Transformer Meets Tracker: Exploiting Temporal Context for Robust Visual Tracking (2021, methodology): “Precision and success plots on the LaSOT test set [13].”
- AiATrack: Attention in Attention for Transformer Visual Tracking (2022, methodology): “Tracker Source LaSOT [17] TrackingNet [45] GOT-10k [25] AUC PNorm P AUC PNorm P AO SR0.”
- Energy-Based Models for Deep Probabilistic Regression (2020, methodology): “We use the training splits of the TrackingNet [18], LaSOT [4], GOT10k [10], and COCO datasets for our training.”
- SiamRPN++: Evolution of Siamese Visual Tracking With Very Deep Networks (2018, methodology): “ test set in terms of success, precision, and normalized precision. LaSOT Dataset. To further validate the proposed framework on a larger and more challenging dataset, we conduct experiments on LaSOT [10]. The LaSOT dataset provides a large-scale, high-quality dense annotations with 1,400 videos in ”
- Autoregressive Visual Tracking (2023, methodology): “The training set consists of GOT10k [31] (we removed 1k sequences in GOT-10k train split according to [61]), LaSOT [19] and TrackingNet [47].”
- Towards More Flexible and Accurate Object Tracking with Natural Language: Algorithms and Benchmark (2021, methodology): “5M 10 fps 6 Train/Eval ✓ LaSOT [15] 1,400 1000 2506 11397 3.”
- Know Your Surroundings: Exploiting Scene Information for Object Tracking (2020, methodology): “on for Object Tracking 11 We use S= 8 dimensional state vectors to encode the scene information. The threshold in the predictor Pis set to 0:05. We use the training splits of TrackingNet [36], LaSOT [13], and GOT-10k [23] datasets to train our network. Within a sequence, we perturb the target positi”
- SwinTrack: A Simple and Strong Baseline for Transformer Tracking (2021, methodology): “LaSOT [15] is a large-scale benchmark containing 280 test sequences.”
- Graph Attention Tracking (2020, neutral): “We evaluate our SiamGAT on several challenge benchmarks, including GOT-10k [14], OTB-100 [31], UAV123 [21] and LaSOT [7].”
- Ocean: Object-aware Anchor-free Tracking (2020, methodology): “Compared to these benchmark datasets, LaSOT [8] has longer sequences, with an average of 2,500 frames per sequence.”
- Deformable Siamese Attention Networks for Visual Object Tracking (2020, methodology): “We use ResNet-50, pre-trained on ImageNet [9], as the backbone, and the whole networks are then fine-tuned on the training sets of COCO [27], YouTube-VOS [39], LaSOT [10] and TrackingNet [30].”
- Learning Target Candidate Association to Keep Track of What Not to Track (2021, methodology): “Success plots, showing OPT , on LaSOT [21] and LaSOTExtSub [20].”