Convolutional Features for Correlation Filter Based Visual Tracking

By Martin Danelljan, Gustav Häger, Fahad Shahbaz Khan, Michael Felsberg. 2015 IEEE International Conference on Computer Vision Workshop (ICCVW), 2015

925 citations recorded by the source corpus.

Visual object tracking is a challenging computer vision problem with numerous real-world applications. This paper investigates the impact of convolutional features for the visual tracking problem. We propose to use activations from the convolutional layer of a CNN in discriminative correlation filter based tracking frameworks. These activations have several advantages compared to the standard deep features (fully connected layers). Firstly, they miti-gate the need of task specific fine-tuning. Secondly, they contain structural information crucial for the tracking problem. Lastly, these activations have low dimensionality. We perform comprehensive experiments on three benchmark datasets: OTB, ALOV300++ and the recently introduced VOT2015. Surprisingly, different to image classification, our results suggest that activations from the first layer provide superior tracking performance compared to the deeper layers. Our results further show that the convolutional features provide improved results compared to standard hand-crafted features. Finally, results comparable to state-of-the-art trackers are obtained on all three benchmark datasets.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/311bc4e48838d8e5ef619df3ce0bc598aba788a1

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

Citations in the corpus

Multi-task Correlation Particle Filter for Robust Object Tracking (2017, methodology): “We evaluate the proposed MCPF algorithm on the Temple Color dataset [22] with 16 trackers in [22] and other 9 state-of-the-art tracking methods using their shared source codes, including MUSTer [18], SRDCF [10], CF2 [25], HDT [29], DSST [9], Staple [2], DeepSRDCF [7], SRDCFdecon [8], and C-COT [11].”
Learning Attentions: Residual Attentional Siamese Network for High Performance Online Visual Tracking (2018, methodology): “For example, CF2 [35] and DeepSRDCF [11] concatenated features from different layers of a pretrained CNN such as VGG [40] into correlation filter.”
Action-Decision Networks for Visual Tracking with Deep Reinforcement Learning (2017, methodology): “ADNet-fast, the fast version of ADNet, has a 3% 2We evaluated the GOTURN [12] on the OTB dataset using the author’s code. performance degradation but runs in real-time (15 fps) and shows performance comparable with those of other CNNbased trackers such as DeepSRDCF [8] and HDT [25].”
Attentional Correlation Filter Network for Adaptive Visual Tracking (2017, methodology): “Deep learning + correlation filter trackers: Correlation filter based trackers show state-of-the-art performance when rich features such as deep convolution features are utilised [6, 8, 26, 30].”
Learning Background-Aware Correlation Filters for Visual Tracking (2017, methodology): “The highest robustness (0.82) belongs to CCOT, followed by DeepSRDCF (1.05) and SRDCF (1.24).”
Adaptive Decontamination of the Training Set: A Unified Formulation for Discriminative Visual Tracking (2016, background): “(1) covers a variety of learning approaches, including support vector machines (SVMs) [11, 25, 29] and discriminative correlation filters (DCFs) [3, 4, 5, 13].”
Deep Learning for Visual Tracking: A Comprehensive Survey (2019, methodology): “Thus, we focus merely on extensive state-of-the-art DL-based trackers, namely: HCFT [51], DeepSRDCF [52], FCNT [53], CNN-”
Learning Aberrance Repressed Correlation Filters for Real-Time UAV Tracking (2019, methodology): “Besides hand-crafted features used in [7, 13, 15, 18], the application of deep features is also investigated for more precise and comprehensive object appearance representation [6,12,19].”
Fully-Convolutional Siamese Networks for Object Tracking (2016, methodology): “correlation filters) using the network’s internal representation as features [5,6] or perform SGD (stochastic gradient descent) to fine-tune multiple layers of the network [7–9].”
Staple: Complementary Learners for Real-Time Tracking (2015, methodology): “Two of the three methods that outperform Staple rely on CNN features, and all run significantly slower: MDNet [11] claims 1 fps, DeepSRDCF [2] <1 fps and SRDCF [4] 5 fps.”
AutoTrack: Towards High-Performance Visual Tracking for UAV With Automatic Spatio-Temporal Regularization (2020, background): “, deep feature representation [18, 34, 17, 20], reinforcement learning [16], residual learning [35] and adversarial learning [36].”
End-to-End Representation Learning for Correlation Filter Based Tracking (2017, methodology): “[21, 7, 9, 31]), which have shown that CNNs and CFs are complementary and their combination results in improved performance.”
Unveiling the Power of Deep Tracking (2018, background): “Other DCF methods focus on integrating convolutional features from a fixed pre-trained deep network [6, 7, 9, 11, 23, 27, 38].”
A Twofold Siamese Network for Real-Time Object Tracking (2018, methodology): “Therefore, we also report the performance of some of the best nonreal-time trackers as a reference, including MDNet[29], DeepSRDCF[7], SRDCF[8] and EBT[40].”
Target-Aware Deep Tracking (2019, neutral): “DeepSRDCF [8] 0.”
Visual Tracking via Adaptive Spatially-Regularized Correlation Filters (2019, methodology): “We compare our tracker against recent state-of-the-art trackers including ECO [8], MDNet [30], LSART [31], C-COT [13], DaSiamRPN [39], SiamRPN [21], DeepSRD-CF [10], ACT [5], BACF [15], StructSiam [38], CF2 [29], SRDCF [11], SiamFC [2], Staple [1], CFNet [1] and KCF [18].”
Large Margin Object Tracking with Circulant Feature Maps (2017, methodology): “It is compared with 9 up-to-date CNNs based trackers including C-COT [8], DeepSRDCF[6], HCF[21], HDT[23], STCT[28], CNN-SVM[15], SINT[25], FCNT[27] and SiameseFC[3].”
VITAL: VIsual Tracking via Adversarial Learning (2018, methodology): “We compare VITAL with 29 trackers from the OTB-2013 benchmark [59] and other 28 stateof-the-art trackers including DSST [10], KCF [22], TGPR [16], MEEM [63], RPT [34], LCT [37], MUSTer [25], HCFT [36], FCNT [57], SRDCF [12], CNN-SVM [24], DeepSRDCF [11], DAT [41], Staple [3], SRDCFdecon [13], CCOT [”