Siamese Cascaded Region Proposal Networks for Real-Time Visual Tracking

By Heng Fan, Haibin Ling. Computer Vision and Pattern Recognition, 2018

408 citations recorded by the source corpus.

Recently, the region proposal networks (RPN) have been combined with the Siamese network for tracking, and shown excellent accuracy with high efficiency. Nevertheless, previously proposed one-stage Siamese-RPN trackers degenerate in presence of similar distractors and large scale variation. Addressing these issues, we propose a multi-stage tracking framework, Siamese Cascaded RPN (C-RPN), which consists of a sequence of RPNs cascaded from deep high-level to shallow low-level layers in a Siamese network. Compared to previous solutions, C-RPN has several advantages: (1) Each RPN is trained using the outputs of RPN in the previous stage. Such process stimulates hard negative sampling, resulting in more balanced training samples. Consequently, the RPNs are sequentially more discriminative in distinguishing difficult background (i.e.,, similar distractors). (2) Multi-level features are fully leveraged through a novel feature transfer block (FTB) for each RPN, further improving the discriminability of C-RPN using both high-level semantic and low-level spatial information. (3) With multiple steps of regressions, C-RPN progressively refines the location and shape of the target in each RPN with adjusted anchor boxes in the previous stage, which makes localization more accurate. C-RPN is trained end-to-end with the multi-task loss function. In inference, C-RPN is deployed as it is, without any temporal adaption, for real-time tracking. In extensive experiments on OTB-2013, OTB-2015, VOT-2016, VOT-2017, LaSOT and TrackingNet, C-RPN consistently achieves state-of-the-art results and runs in real-time.

Subfields: Computer Vision and Pattern Recognition, Global and Planetary Change, Safety, Risk, Reliability and Quality

View Paper: https://www.semanticscholar.org/paper/f98be9a91dbf00b52a494720bd36be9c73a1210e

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

In Global and Planetary Change

In Safety, Risk, Reliability and Quality

Citations in the corpus

Graph Attention Tracking (2020, positive): “To tackle data imbalance, C-RPN [8] proposes to cascade a sequence of RPNs from deep high-level to shallow low-level layers in a Siamese network.”
Siamese Box Adaptive Network for Visual Tracking (2020, methodology): “We evaluate our tracker on the test set consisting of 280 videos with trackers including DiMP [2], ATOM [5], SiamRPN++ [20], C-RPN [10], MDNet [28], VITAL [35], StructSiam [48], DSiam[11], ECO [6].”
Deformable Siamese Attention Networks for Visual Object Tracking (2020, background): “Our work is related to that of [11] where a C-RPN was developed to progressively reﬁne the location of target with a sequence of RPNs, but we design a new module that only reﬁnes a single output region, which is particularly lightweight and can be integrated into very deep Siamese networks.”
LaSOT: A High-quality Large-scale Single Object Tracking Benchmark (2020, background): “Owing to balanced efficiency and accuracy, deep Siamese tracking has been extended by many later works [27, 36, 49, 51, 78, 80, 95].”
Ocean: Object-aware Anchor-free Tracking (2020, methodology): “CRPN ECO STRCF LADCF ATOM SRCNN SiamRPN++ DiMP Ocean Ocean [9] [5] [23] [45] [4] [39] [21] [2] offline online EAO ↑ 0.”
Deep Learning for Visual Tracking: A Comprehensive Survey (2019, methodology): “To enhance efficiency and accuracy, the target regions are estimated by classification scores and optimized regression/matching functions [56], [57], [134], [145], [146], [156], [157], [160], [163], [164], [167], [179].”
Transformer Meets Tracker: Exploiting Temporal Context for Robust Visual Tracking (2021, methodology): “The precision and success plots of the state-of-theart methods are shown in Figure 7, where the recently proposed C-RPN [14], SiamRPN++ [27], ATOM [7], DiMP-50 [3], and PrDiMP-50 [9] are included for comparison.”
Siam R-CNN: Visual Tracking by Re-Detection (2019, methodology): “Recent tracking approaches improve upon SiamRPN, making it distractor aware (DaSiamRPN [82]), adding a cascade (C-RPN [19]), producing masks (SiamMask [66]), using deeper architectures (SiamRPN+ [79] and SiamRPN++ [37]) and maintaining a set of diverse templates (THOR [58]).”
Cascade RPN: Delving into High-Quality Region Proposal Network with Adaptive Convolution (2019, methodology): “To alleviate this problem, recent advanced methods [12, 37] use deformable convolution [10] to perform spatial transformations on the features as shown in Figure 2c and 2d and expect transformed features to align to the change in anchor geometry.”
SwinTrack: A Simple and Strong Baseline for Transformer Tracking (2021, background): “LaSOT.”
Learning the Model Update for Siamese Trackers (2019, methodology): “LaSOT is a much larger and more challenging dataset including long-term sequences, Following recent works that use this dataset [26, 14], we report results on protocol II, i.”
SiamCAR: Siamese Fully Convolutional Classification and Regression for Visual Tracking (2019, background): “Current popular visual tracking methods [1, 21, 14, 35, 20, 42, 11, 33] revolve around the Siamese network based architectures.”