Prítel, Pavol · Brno University of Technology Digital Library (Brno University of Technology) · 2014
The thesis deals with detection of moving objects in image sequence. The target of this thesis is a design of object detection algorithm and its application in vehicle counting. The thesis is divided into three parts. The first part of this thesis includes theoretical analysis of given problem. The second part is related to the design and implementation of the algorithm. The third part deals with evaluation of results of the implemented algorithm.
Cited by 0
Shengrong Gong, Chunping Liu, Yi Ji et al. · Modeling and optimization in science and technologies · 2018
Cited by 8
Ezzat, Ahmed · 2021
Cited by 0Open access
Chang, Chu-Yin, English, James D., Tardella, Neil M. · NASA STI Repository (National Aeronautics and Space Administration)
This invention describes a method for identifying and tracking an object from two-dimensional data pictorially representing said object by an object-tracking system through processing said two-dimensional data using at least one tracker-identifier belonging to the object-tracking system for providing an output signal containing: a) a type of the object, and/or b) a position or an orientation of the object in three-dimensions, and/or c) an articulation or a shape change of said object in said three dimensions.
Cited by 0Open access
Fei Chen, Xiaodong Wang, Yunxiang Zhao et al. · Computer Vision and Image Understanding · 2022
Cited by 129
Sudhir Kumar, Sandeep Kumar Singh · 2023
In the present world, the evolution of technology takes place in terms of developing new things to solve some real-world problems. Almost all fields are creative in the sense of creating new things. Computer vision is one of the fields that give the technology that can help with different real-time problems. This field offers some useful applications over the last few decades. Different technologies like Autonomous cars, Traffic flow analysis, X-ray analysis, etc. make this field interesting for researchers. Object detection is one of the applications of Computer vision. Object detection and tracking enable the machine to detect the object and track the object through the entire scene. In this paper, a supervised learning approach for object detection is presented which consists of a detector, a Neural network, Images, and Groundtruths. The presented model successfully tracks the object and the dataset used for the performance evaluation is Online Object Tracking Benchmark (OOTB), Unmanned Aerial Vehicle (UAV), and Visual Object Tracking Benchmark (VOT) dataset.
Cited by 1
Severin Stalder · Repository for Publications and Research Data (ETH Zurich) · 2012
Cited by 0Open access
Mengmeng Wang, X. Kong, Guojiang Shen · Advances in computer vision and pattern recognition · 2026
Cited by 0
Chih-Ching 林智慶;Lin · 2017
碩士
Cited by 0
Jack Hong, Shilin Yan, Zehao Xiao et al. · ArXiv.org · 2025
In this work, we propose a progressive scaling training strategy for visual object tracking, systematically analyzing the influence of training data volume, model size, and input resolution on tracking performance. Our empirical study reveals that while scaling each factor leads to significant improvements in tracking accuracy, naive training suffers from suboptimal optimization and limited iterative refinement. To address this issue, we introduce DT-Training, a progressive scaling framework that integrates small teacher transfer and dual-branch alignment to maximize model potential. The resulting scaled tracker consistently outperforms state-of-the-art methods across multiple benchmarks, demonstrating strong generalization and transferability of the proposed method. Furthermore, we validate the broader applicability of our approach to additional tasks, underscoring its versatility beyond tracking.
Cited by 0Open access
Kratochvíla, Lukáš · Brno University of Technology Digital Library (Brno University of Technology) · 2019
Sledování obecného objektu na zařízení s omezenými prostředky v reálném čase je obtížné. Mnoho algoritmů věnujících se této problematice již existuje. V této práci se s nimi seznámíme. Různé přístupy k této problematice jsou diskutovány včetně hlubokého učení. Představeny jsou reprezentace objektu, datasety i metriky pro vyhodnocování. Mnoho sledovacích algorimů je představeno, osm z nich je implementováno a vyhodnoceno na VOT datasetu.
Cited by 0
Navarro Comes, Albert · RECERCAT (Consorci de Serveis Universitaris de Catalunya)
The presence of robots in humans’ daily life is a growing phenomenon. Autonomous navigation is a part of this progression. In this study robust tracking of different objects simultaneously is investigated.\nA software capable of tracking different objects, mainly humans and robots, is built using a Kinect camera situated in a moving platform. Computer vision technology is used to provide a tool to improve the human and robot interaction. Choosing a sensor like the recently released Kinect camera, colour and depth images are combined to obtain a differential gain in the tracking implementation. The sensor has been mounted in a moving robot able to provide the position of the robot at any moment from the sensor on its wheels. This information is very valuable to transform the data from different frames and positions to the same reference system.\nThe use given to the Kinect concerns mainly the tracking of one or two humans from a fixed position. This work shows that it can also be used to track humans and much smaller objects simultaneously when the camera is mounted on a mobile robot.\nThe Point Cloud Library is the framework used to deal with the data obtained from the sensor. The use of PCL libraries for filtering, segmentation and clustering results in the detection of the object. Then the descriptor creation and comparison with the ones obtained in previous frames will come up with the correlation part. A training process for the recognition of humans and robots is performed beforehand. The descriptor used is the Viewpoint Feature Histogram, which maintains invariance to distance and size.\nThe main limitation imposed is that the solution implemented should be integrated in the Mobotware framework. Thus, it has been created as a plug-in running directly in the framework. The Kinect introduce limitations such as small range of view and impossibility of being used outdoors.\nThe application is validated with tests and the results obtained are presented. The recognition of the objects trained has proved to be successful and the time of computation, even though not ideal, is enough to see how the scene varies over time. Enough to track humans and SMR moving at a normal speed. In all cases the system detects and tracks the objects and keeps learning from the detections over time. Even though it tracks well, when movement is added to the camera the tracking process is less effective, mainly due to the apparition of false detections.
Cited by 0Open access
程, 子一 · Kyushu University Institutional Repository (QIR) (Kyushu University) · 2023
Visual object tracking (VOT) is one of the most widely studied computer vision approaches that can produce the trajectory of a moving object from a sequence of frames.VOT has many applications, such as navigation for robots, intelligent video surveillance, smart logistics, robotics for manufacturing, etc.Most VOT models are based on deep neural networks (DNNs) because of their powerful feature extraction capability.However, most of the current VOT research has focused on the design of model architectures that is limited to exploring the full potential of the VOT model.In this thesis, we take the following three aspects to build a more powerful and reliable visual tracking scheme: Data Augmentation is a data pre-processing technique commonly used with deep learning models to mitigate overfitting during training.In the context of visual object tracking, many state-of-the-art trackers employ an online updating strategy, wherein a classifier is dynamically trained in real-time during testing to accurately locate the target.The online updating of the object model using samples from previous frames plays a pivotal role in achieving precise visual object tracking.Exploring how to effectively leverage these samples, such as through data augmentation, offers a novel approach to improve tracking performance.By applying data augmentation techniques to the collected samples, we can introduce variations and diversify the training data, thereby enhancing the model's ability to generalize to different visual conditions.This, in turn, enables the tracker to better adapt to the inherent challenges encountered in real-world
Cited by 0Open access
Geng Zhang, Zejian Yuan, Nanning Zheng et al. · Lecture notes in computer science · 2010
Cited by 40
Rui Yao, Guosheng Lin, Chunhua Shen et al. · IEEE Transactions on Circuits and Systems for Video Technology · 2018
In this paper, we propose a semantics-aware visual object tracking method, which introduces semantics into the tracking procedure and extends the model of an object with explicit semantics prior to enhancing the robustness of three key aspects of the tracking framework, i.e., appearance model, search scheme, and scale adaptation. We first present a semantic object proposal generation method for video sequences to generate high-quality category-oriented object proposals. Then, a hybrid semantics-aware tracking algorithm with semantic compatibility is proposed. This algorithm takes full advantages of globally sparse semantic object proposal prediction and locally dense prediction with a template model and semantic distractor-aware color appearance model. Furthermore, we propose to exploit semantics to localize object accurately via an energy minimization framework-based scale adaptation method, which jointly integrates dense location prior, instance-specific color, and category-specific semantic information. Extensive experiments are conducted on two widely used benchmarks, and the results demonstrate that our method achieves the state-of-the-art performance.
Cited by 42Open access
Liangtao Shi, Bineng Zhong, Qihua Liang et al. · arXiv (Cornell University) · 2024
How to effectively exploit spatio-temporal information is crucial to capture target appearance changes in visual tracking. However, most deep learning-based trackers mainly focus on designing a complicated appearance model or template updating strategy, while lacking the exploitation of context between consecutive frames and thus entailing the \textit{when-and-how-to-update} dilemma. To address these issues, we propose a novel explicit visual prompts framework for visual tracking, dubbed \textbf{EVPTrack}. Specifically, we utilize spatio-temporal tokens to propagate information between consecutive frames without focusing on updating templates. As a result, we cannot only alleviate the challenge of \textit{when-to-update}, but also avoid the hyper-parameters associated with updating strategies. Then, we utilize the spatio-temporal tokens to generate explicit visual prompts that facilitate inference in the current frame. The prompts are fed into a transformer encoder together with the image tokens without additional processing. Consequently, the efficiency of our model is improved by avoiding \textit{how-to-update}. In addition, we consider multi-scale information as explicit visual prompts, providing multiscale template features to enhance the EVPTrack's ability to handle target scale changes. Extensive experimental results on six benchmarks (i.e., LaSOT, LaSOT\rm $_{ext}$, GOT-10k, UAV123, TrackingNet, and TNL2K.) validate that our EVPTrack can achieve competitive performance at a real-time speed by effectively exploiting both spatio-temporal and multi-scale information. Code and models are available at https://github.com/GXNU-ZhongLab/EVPTrack.
Cited by 1Open access
Shih-Fang Chen, Juncheng Chen, I‐Hong Jhuo et al. · arXiv (Cornell University) · 2024
Learning a discriminative model that distinguishes the specified target from surrounding distractors across frames is essential for generic object tracking (GOT). Dynamic adaptation of target representation against distractors remains challenging because prevailing trackers exhibit limited discriminative capability. To address this issue, we present a new visual prompting mechanism for generic object tracking, termed PiVOT. PiVOT introduces mechanisms that leverage the pretrained foundation model (CLIP) to automatically generate and refine visual prompts online, thereby enabling the tracker to suppress distractors through contrastive guidance. To transfer contrastive knowledge from the foundation model to the tracker, PiVOT automatically propagates this knowledge online and dynamically generates and updates visual prompts. Specifically, it proposes a prompt initialization mechanism that produces an initial visual prompt highlighting potential target locations. The foundation model is then used to refine the prompt based on appearance similarities between candidate objects and reference templates across potential targets. After refinement, the visual prompt better highlights potential target locations and reduces irrelevant prompt information. With the proposed prompting mechanism, the tracker can generate instance-aware feature maps guided by the visual prompts, which are incrementally and automatically updated during tracking, thereby effectively suppressing distractors. Extensive experiments across multiple benchmarks indicate that PiVOT, with the proposed prompting mechanism, can suppress distracting objects and improve tracking performance.
Cited by 0Open access
Shih-Fang Chen, Jun-Cheng Chen, I‐Hong Jhuo et al. · IEEE Transactions on Multimedia · 2025
Learning a discriminative model to distinguish a target from its surrounding distractors is essential to generic visual object tracking. Dynamic target representation adaptation against distractors is challenging due to the limited discriminative capabilities of prevailing trackers. We present a new visual Prompting mechanism for generic Visual Object Tracking (PiVOT) to address this issue. PiVOT proposes a prompt generation network with the pre-trained foundation model CLIP to automatically generate and refine visual prompts, enabling the transfer of foundation model knowledge for tracking. While CLIP offers broad category-level knowledge, the tracker, trained on instance-specific data, excels at recognizing unique object instances. Thus, PiVOT first compiles a visual prompt highlighting potential target locations. To transfer the knowledge of CLIP to the tracker, PiVOT leverages CLIP to refine the visual prompt based on the similarities between candidate objects and the reference templates across potential targets. Once the visual prompt is refined, it can better highlight potential target locations, thereby reducing irrelevant prompt information. With the proposed prompting mechanism, the tracker can generate improved instance-aware feature maps through the guidance of the visual prompt, thus effectively reducing distractors. The proposed method does not involve CLIP during training, thereby keeping the same training complexity and preserving the generalization capability of the pretrained foundation model. Extensive experiments across multiple benchmarks indicate that PiVOT, using the proposed prompting method can suppress distracting objects and enhance the tracker.
Cited by 2
Shih-Fang Chen, Juncheng Chen, I‐Hong Jhuo et al. · 2024
Learning a discriminative model to distinguish a target from its surrounding distractors is essential to generic visual object tracking. Dynamic target representation adaptation against distractors is challenging due to the limited discriminative capabilities of prevailing trackers. We present a new visual Prompting mechanism for generic Visual Object Tracking (PiVOT) to address this issue. PiVOT proposes a prompt generation network with the pre-trained foundation model CLIP to automatically generate and refine visual prompts, enabling the transfer of foundation model knowledge for tracking. While CLIP offers broad category-level knowledge, the tracker, trained on instance-specific data, excels at recognizing unique object instances. Thus, PiVOT first compiles a visual prompt highlighting potential target locations. To transfer the knowledge of CLIP to the tracker, PiVOT leverages CLIP to refine the visual prompt based on the similarities between candidate objects and the reference templates across potential targets. Once the visual prompt is refined, it can better highlight potential target locations, thereby reducing irrelevant prompt information. With the proposed prompting mechanism, the tracker can generate improved instance-aware feature maps through the guidance of the visual prompt, thus effectively reducing distractors. The proposed method does not involve CLIP during training, thereby keeping the same training complexity and preserving the generalization capability of the pretrained foundation model. Extensive experiments across multiple benchmarks indicate that PiVOT, using the proposed prompting method can suppress distracting objects and enhance the tracker. The code is publicly available.
Cited by 1Open access
Shih-Fang Chen, Juncheng Chen, I‐Hong Jhuo et al. · 2024
Learning a discriminative model to distinguish a target from its surrounding distractors is essential to generic visual object tracking. Dynamic target representation adaptation against distractors is challenging due to the limited discriminative capabilities of prevailing trackers. We present a new visual Prompting mechanism for generic Visual Object Tracking (PiVOT) to address this issue. PiVOT proposes a prompt generation network with the pre-trained foundation model CLIP to automatically generate and refine visual prompts, enabling the transfer of foundation model knowledge for tracking. While CLIP offers broad category-level knowledge, the tracker, trained on instance-specific data, excels at recognizing unique object instances. Thus, PiVOT first compiles a visual prompt highlighting potential target locations. To transfer the knowledge of CLIP to the tracker, PiVOT leverages CLIP to refine the visual prompt based on the similarities between candidate objects and the reference templates across potential targets. Once the visual prompt is refined, it can better highlight potential target locations, thereby reducing irrelevant prompt information. With the proposed prompting mechanism, the tracker can generate improved instance-aware feature maps through the guidance of the visual prompt, thus effectively reducing distractors. The proposed method does not involve CLIP during training, thereby keeping the same training complexity and preserving the generalization capability of the pretrained foundation model. Extensive experiments across multiple benchmarks indicate that PiVOT, using the proposed prompting method can suppress distracting objects and enhance the tracker. The code is publicly available1.
Cited by 1Open access