Luka Čehovin Zajc
A5070543072- Video Surveillance and Tracking Methods
- Human Pose and Action Recognition
- Advanced Image and Video Retrieval Techniques
- Advanced Vision and Imaging
- Infrared Target Detection Methodologies
- Fire Detection and Safety Systems
- Impact of Light on Environment and Health
- Face recognition and analysis
- Visual Attention and Saliency Detection
- Remote Sensing in Agriculture
- Video Analysis and Summarization
- Advanced Neural Network Applications
- Air Quality Monitoring and Forecasting
- Advanced Measurement and Detection Methods
- 3D Surveying and Cultural Heritage
- Spectroscopy and Chemometric Analyses
- Teaching and Learning Programming
- Robotics and Automated Systems
- Image Enhancement Techniques
- Smart Agriculture and AI
- Engineering Education and Technology
- Soft Robotics and Applications
- Digital Rights Management and Security
- Data Visualization and Analytics
- Mechatronics Education and Applications
University of Ljubljana
2015-2024
Short-term tracking is an open and challenging problem for which discriminative correlation filters (DCF) have shown excellent performance.We introduce the channel spatial reliability concepts to DCF provide a novel learning algorithm its efficient seamless integration in filter update process. The map adjusts support part of object suitable tracking. This both allows enlarge search region improves non-rectangular objects. Reliability scores reflect channel-wise quality learned are used as...
The Visual Object Tracking challenge 2015, VOT2015, aims at comparing short-term single-object visual trackers that do not apply pre-learned models of object appearance. Results 62 are presented. number tested makes VOT 2015 the largest benchmark on tracking to date. For each participating tracker, a short description is provided in appendix. Features VOT2015 go beyond its VOT2014 predecessor are: (i) new dataset twice as large with full annotation targets by rotated bounding boxes and...
This paper addresses the problem of single-target tracker performance evaluation. We consider measures, dataset and evaluation system to be most important components propose requirements for each them. The are basis a new methodology that aims at simple easily interpretable comparison. ranking-based equivalence in terms statistical significance practical differences. A fully-annotated with per-frame annotations several visual attributes is introduced. diversity its properties maximized novel...
The Visual Object Tracking challenge VOT2017 is the fifth annual tracker benchmarking activity organized by VOT initiative. Results of 51 trackers are presented; many state-of-the-art published at major computer vision conferences or journals in recent years. evaluation included standard and other popular methodologies a new "real-time" experiment simulating situation where processes images as if provided continuously running sensor. Performance tested typically far exceeds baselines. source...
The Visual Object Tracking challenge VOT2019 is the seventh annual tracker benchmarking activity organized by VOT initiative. Results of 81 trackers are presented; many state-of-the-art published at major computer vision conferences or in journals recent years. evaluation included standard and other popular methodologies for short-term tracking analysis as well methodology long-term analysis. was composed five challenges focusing on different domains: (i) VOTST2019 focused RGB, (ii)...
Visual tracking has attracted a significant attention in the last few decades. The recent surge number of publications on tracking-related problems have made it almost impossible to follow developments field. One reasons is that there lack commonly accepted annotated data-sets and standardized evaluation protocols would allow objective comparison different methods. To address this issue, Object Tracking (VOT) workshop was organized conjunction with ICCV2013. Researchers from academia as well...
This paper addresses the problem of tracking objects which undergo rapid and significant appearance changes. We propose a novel coupled-layer visual model that combines target's global local by interlacing two layers. The layer in this is set patches geometrically constrain changes appearance. probabilistically adapts to geometric deformation, while its structure updated removing adding patches. addition these constrained models properties, such as color, shape, apparent motion. properties...
The problem of visual tracking evaluation is sporting a large variety performance measures, and largely suffers from lack consensus about which measures should be used in experiments. This makes the cross-paper tracker comparison difficult. Furthermore, as some may less effective than others, results skewed or biased towards particular aspects. In this paper we revisit popular visualizations analyze them theoretically experimentally. We show that several are equivalent point information they...
Deformable parts models show a great potential in tracking by principally addressing nonrigid object deformations and self occlusions, but according to recent benchmarks, they often lag behind the holistic approaches. The reason is that potentially large number of degrees freedom have be estimated for localization simplifications constellation topology are assumed make inference tractable. We present new formulation model with correlation filters treats geometric visual constraints within...
The Thermal Infrared Visual Object Tracking challenge 2015, VOT-TIR2015, aims at comparing short-term single-object visual trackers that work on thermal infrared (TIR) sequences and do not apply pre-learned models of object appearance. VOT-TIR2015 is the first benchmark tracking in TIR sequences. Results 24 are presented. For each participating tracker, a short description provided appendix. based VOT2013 challenge, but introduces following novelties: (i) newly collected LTIR (Link -- ping...
The Visual Object Tracking challenge VOT2021 is the ninth annual tracker benchmarking activity organized by VOT initiative. Results of 71 trackers are presented; many state-of-the-art published at major computer vision conferences or in journals recent years. was composed four sub-challenges focusing on different tracking domains: (i) VOT-ST2021 focused short-term RGB, (ii) VOT-RT2021 "real-time" (iii) VOT-LT2021 long-term tracking, namely coping with target disappearance and reappearance...
Crop classification is an important task in remote sensing with many applications, such as estimating yields, detecting crop diseases and pests, ensuring food security. In this study, we combined knowledge from sensing, machine learning, agriculture to investigate the application of transfer learning a transformer model for variable length satellite image time series (SITS). The objective was produce map agricultural land, reduce required interventions, limit in-field visits. Specifically,...
This paper addresses the problem of tracking objects which undergo rapid and significant appearance changes. We propose a novel coupled-layer visual model that combines target's global local appearance. The layer in this is set patches geometrically constrain changes probabilistically adapts to geometric deformation, while its structure updated by removing adding patches. addition constrained models properties such as color, shape apparent motion. are during using stable from layer. By...
The problem of visual tracking evaluation is sporting an abundance performance measures, which are used by various authors, and largely suffers from lack consensus about measures should be preferred. This hampering the cross-paper tracker comparison faster advancement field. In this paper we provide overview popular visualizations their critical theoretical experimental analysis. We show that several equivalent point information they for and, crucially, some more brittle than others. Based...
We propose a new long-term tracking performance evaluation methodology and present challenging dataset of carefully selected sequences with many target disappearances. perform an extensive six nine short-term state-of-the-art trackers, using measures, suitable for evaluating - precision, recall F-score. The shows that good model update strategy the capability image-wide re-detection are critical performance. integrated in VOT toolkit to automate experimental analysis benchmarking facilitate...
In the paper, we present an empirical evaluation of five feature selection methods: ReliefF, random forest selector, sequential forward selection, backward and Gini index. Among evaluated methods, f
Deformable part models exhibit excellent performance in tracking non-rigidly deforming targets, but are usually outperformed by holistic when the target does not deform or presence of uncertain visual data. The reason is that part-based require estimation a larger number parameters compared to and since updating process self-supervised, errors parameter amplified with time, leading faster accuracy reduction than models. On other hand, robustness trackers generally greater trackers. We...