Given the pervasiveness of web tracking practices on Internet, many countries are developing and enforcing new privacy regulations to ensure rights their citizens. However, discovering websites that do not comply with those is becoming very challenging, given dynamic nature or use obfuscation techniques. This work presents ePrivo, a online service can help Internet users, website owners, regula...

In this work we develop a novel temporal decomposition scheme for solving data association in multisensormultitarget tracking problems. Given a set of noisy measurements from any number of sensors the data association problem is to determine which measurements originate from which targets. The problem is traditionally posed as an N -dimensional assignment problem, which is NP-hard for N ≥ 3 dim...

For the ALICE High Level Trigger a fast tracking algorithm was developed by Sergey Gorbunov based on the Cellular Automaton method and the Kalman filter [1], that is currently installed in the HLT. For an efficient handling of upcoming lead-lead collisions in 2010 with a tremendous increase of clusters and tracks, possibilities for a better usage of parallelism and many core hardware were analy...

Many universities are currently using a Course Management System (CMS) to organize course materials and conduct online learning activities. However, most commercial or open source CMS software does not include comprehensive access tracking and log analysis capabilities. In this paper, we propose and implement a CMS log analysis tool, called Moodog, to track students’ online learning activities....

We introduce Bayesian online learning for real time parameter adaptation on a tempo tracking task. We employ a variational extension of the Expectation-Maximization algorithm for online parameter estimation. Simulation results on a real dataset indicate that online adaptation has the potential of capturing performer specific features in real time.

To achieve effective visual tracking, a robust feature representation composed of two separate components (i.e., feature learning and selection) for an object is one of the key issues. Typically, a common assumption used in visual tracking is that the raw video sequences are clear, while real-world data is with significant noise and irrelevant patterns. Consequently, the learned features may be...

Object tracking is one of the most important tasks in many applications of computer vision. Many tracking methods use a fixed set of features ignoring that appearance of a target object may change drastically due to intrinsic and extrinsic factors. The ability to dynamically identify discriminative features would help in handling the appearance variability by improving tracking performance. The...

Multi-person articulated pose tracking in complex unconstrained videos is an important and challenging problem. In this paper, going along the road of top-down approaches, we propose a decent and efficient pose tracker based on pose flows. First, we design an online optimization framework to build association of cross-frame poses and form pose flows. Second, a novel pose flow non maximum suppre...

This paper evaluates the WiSARD weightless model as a classification system on the problem of tracking multiple objects in realtime. Exploring the structure of this model, the proposed solution applies a re-learning stage in order to avoid interferences caused by background noise or variations in the target shape. Once the tracker finds a target at the first time, it applies only local searches...

We present more evaluation results in this document. Tracking Speed. Table 1 shows the statistics of the tracking speed of each algorithm in OPE running on a PC with Intel i7 3770 CPU (3.4GHz). The speed of L1APG is slower than [4] as we set the parameters of L1APG to be the default ones of MTT, where the canonical size of template is larger than the default one of L1APG. The implementation of ...