Convolutional Neural Networks (CNN) have been demonstrated to achieve state-of-the-art performance in visual object tracking task. However, existing CNN-based trackers usually use holistic target samples to train their networks. Once the target undergoes complicated situations (e.g., occlusion, background clutter, and deformation), the tracking performance degrades badly. In this paper, we propose an adaptive structural convolutional filter model to enhance the robustness of deep regression trackers (named: ASCT). Specifically, we first design a mask set to generate local filters to capture local structures of the target. Meanwhile, we adopt an adaptive weighting fusion strategy for these local filters to adapt to the changes in the target appearance, which can enhance the robustness of the tracker effectively. Besides, we develop an end-to-end trainable network comprising feature extraction, decision making, and model updating modules for effective training. Extensive experimental results on large benchmark datasets demonstrate the effectiveness of the proposed ASCT tracker performs favorably against the state-of-the-art trackers.

Visual object tracking with adaptive structural convolutional network

Recently, deep learning has achieved great success in visual tracking tasks, particularly in single-object tracking. This paper provides a comprehensive review of state-of-the-art single-object tracking algorithms based on deep learning. First, we introduce basic knowledge of deep visual tracking, including fundamental concepts, existing algorithms, and previous reviews. Second, we briefly review existing deep learning methods by categorizing them into data-invariant and data-adaptive methods based on whether they can dynamically change their model parameters or architectures. Then, we conclude with the general components of deep trackers. In this way, we systematically analyze the novelties of several recently proposed deep trackers. Thereafter, popular datasets such as Object Tracking Benchmark (OTB) and Visual Object Tracking (VOT) are discussed, along with the performances of several deep trackers. Finally, based on observations and experimental results, we discuss three different characteristics of deep trackers, i.e., the relationships between their general components, exploration of more effective tracking frameworks, and interpretability of their motion estimation components.

/pdf/advances-in-deep-learning-methods-for-visual-tracking-49twpn0vvl.pdf

Advances in Deep Learning Methods for Visual Tracking: Literature Review and Fundamentals

Siamese networks based visual tracking has recently drawn great attention due to their superior representation and tracking accuracy However, the backbone networks and prediction networks still cannot fully take advantage of features from modern deep networks In this paper, we propose an inverted residual Siamese feature-crossing network (IRSiamese-FCN) which is end-to-end trained off-line with a large amount of image pairs Specifically, the Siamese backbone networks for feature extraction consist of an inverted residual network and a feature-crossing network (FCN) The designed IR architecture is light weighted by combination of depthwise and pointwise convolutions Moreover, non-linearities and linearities are proceeded separately in deep and narrow layers Feature-crossing network is to perform feature-level aggregations, which makes deep and shallow layers complement each other more closely and further improves tracking accuracy We conduct ablation studies and comparison experiments over five large benchmarks The results demonstrate that the proposed tracker can achieve competitive performance

/pdf/inverted-residual-siamese-visual-tracking-with-feature-2cpxah9l6y.pdf

Inverted Residual Siamese Visual Tracking With Feature Crossing Network

Multiple-object tracking is a fundamental computer vision task which is gaining increasing attention due to its academic and commercial potential. Multiple-object detection, recognition and tracking are quite desired in many domains and applications. However, accurate object tracking is very challenging, and things are even more challenging when multiple objects are involved. The main challenges that multiple-object tracking is facing include the similarity and the high density of detected objects, while also occlusions and viewpoint changes can occur as the objects move. In this article, we introduce a real-time multiple-object tracking framework that is based on a modified version of the Deep SORT algorithm. The modification concerns the process of the initialization of the objects, and its rationale is to consider an object as tracked if it is detected in a set of previous frames. The modified Deep SORT is coupled with YOLO detection methods, and a concrete and multi-dimensional analysis of the performance of the framework is performed in the context of real-time multiple tracking of vehicles and pedestrians in various traffic videos from datasets and various real-world footage. The results are quite interesting and highlight that our framework has very good performance and that the improvements on Deep SORT algorithm are functional. Lastly, we show improved detection and execution performance by custom training YOLO on the UA-DETRAC dataset and provide a new vehicle dataset consisting of 7 scenes, 11.025 frames and 25.193 bounding boxes.

Real-time multiple object tracking using deep learning methods

In complex scenes with light changes, deformations, and occlusions, target tracking easily contains a large amount of background color information when building a target color model. Thus, the tracking effect is reduced. To improve the accuracy of the traditional continuously adaptive mean-shift algorithm (CAMShift) in complex scenarios, a target tracking algorithm based on an improved Gaussian mixture model was proposed. Using the Gaussian mixture model, the tracking image was divided into the foreground and background superposition. The histograms of the hue component were respectively established in the foreground and background of the target area. By suppressing the same hue as the background color in the tracking image, the target color model was established. The target position was iteratively obtained by implementing the CAMShift algorithm using the enhanced target color model. The Bhattacharyya distance between the candidate target and the target template was used as basis for updating the target model. Simulation analysis under benchmark data sets and actual monitoring scenarios verified the accuracy of the proposed algorithm. Results show that the distance precision and overlap success rate of the proposed algorithm are 0.88 and 0.625, respectively. The proposed algorithm effectively solves long-term target tracking problems with complex scenes, such as occlusion, background clutters, and illumination variation. This study eliminates the problem of target recognition caused by environmental changes and provides references for real-time monitoring of abnormal traffic conditions.

/pdf/continuously-adaptive-mean-shift-tracking-algorithm-based-on-4kh57rrvqi.pdf

Fei Chen

Papers

Visual object tracking: A survey