PhD Thesiss

2D shape recognition using information theoretic kernels

Bicego M., Martins A.F.T., Murino V., Aguiar P.M.Q., Figueiredo M.A.T.

Proceedings - International Conference on Pattern Recognition

2010

Abstract:

In this paper, a novel approach for contour based 2D shape recognition is proposed, using a class of information theoretic kernels recently introduced. This kind of kernels, based on a non-extensive generalization of the classical Shannon information theory, are defined on probability measures. In the proposed approach, chain code representations are first extracted from the contours; then n-gram statistics are computed and used as input to the information theoretic kernels. We tested different versions of such kernels, using support vector machine and nearest neighbor classifiers. An experimental evaluation on the Chicken pieces dataset shows that the proposed approach significantly outperforms the current state-of-the-art methods.

6D pose estimation of textureless shiny objects using random ferns for bin-picking

Rodrigues J.J., Kim J.-S., Furukawa M., Xavier J., Aguiar P., Kanade T.

IEEE International Conference on Intelligent Robots and Systems

2012

Abstract:

We address the problem of 6D pose estimation of a textureless and shiny object from single-view 2D images, for a bin-picking task. For a textureless object like a mechanical part, conventional visual feature matching usually fails due to the absence of rich texture features. Hierarchical template matching assumes that few templates can cover all object appearances. However, the appearance of a shiny object largely depends on its pose and illumination. Furthermore, in a bin-picking task, we must cope with partial occlusions, shadows, and inter-reflections.

A Benchmark of Visual Storytelling in Social Media

Marcelino G., Semedo D., Mourão A., Blasi S., Mrak M., Magalhães J.

Proceedings of the 2019 on International Conference on Multimedia Retrieval

2019

Abstract:

Media editors in the newsroom are constantly pressed to provide a”like-being there” coverage of live events. Social media provides a disorganised collection of images and videos that media professionals need to grasp before publishing their latest news updated. Automated news visual storyline editing with social media content can be very challenging, as it not only entails the task of finding the right content but also making sure that news content evolves coherently over time. To tackle these issues, this paper proposes a benchmark for assessing social media visual storylines. The SocialStories benchmark, comprised by total of 40 curated stories covering sports and cultural events, provides the experimental setup and introduces novel quantitative metrics to perform a rigorous evaluation of visual storytelling with social media data.

A cautionary view of mobility and connectivity modeling in vehicular ad-hoc networks

Conceicao H., Ferreira M., Barros J.

IEEE Vehicular Technology Conference

2009

Abstract:

Motivated by the recent surge in vehicular ad-hoc network (VANET) research and the promise of high-impact applications such as safety, navigation and infotainment services, we consider the impact of mobility mis-modeling on the design and development of this class of distributed systems. Focusing on urban environments, we use a state-of-the-art car traffic simulator to extract some of the key connectivity metrics relevant for the design of a vehicle-to-vehicle traffic information system. We compare such metrics against those obtained from popular mobility models, such as the random waypoint model and the Manhattan mobility model. Our results reveal striking differences in the connectivity profile of the network, thus casting some doubt on the adequacy of simple mobility models for the development of future VANET protocols.

A Cloud-Optimized Link Layer for Low-Power Wide-Area Networks

Balanuta A.; Pereira N.; Kumar S.; Rowe A.

MobiSys

2020

Abstract:

Conventional wireless communication systems are typically designed assuming a single transmitter-receiver pair for each link. In Low-Power Wide-Area Networks (LP-WANs), this one-to-one design paradigm is often overly pessimistic in terms of link budget because client packets are frequently detected by multiple gateways (i.e. one-to-many). Prior work has shown massive improvement in performance when specialized hardware is used to coherently combine signals at the physical layer. In this paper, we explore the potential of using multiple receivers at the MAC and link layer where these performance gains are often neglected. We present an approach called Opportunistic Packet Recovery (OPR) that targets the most likely corrupt bits across a set of packets that suffered failed CRCs at multiple LoRa LP-WAN base-stations. We see that bit errors are often disjoint across receivers, which aids in collaborative error detection. OPR leverages this to provide increasing gain in error recovery as a function of the number of receiving gateways. Since LP-WAN networks can easily offload packet processing to the cloud, there is ample compute time per packet (order of seconds) to search for bit permutations that would restore packet integrity. Link layer corrections have the advantage of being immediately applicable to the millions of already deployed LP-WAN systems without additional hardware or expensive RF front-ends. We experimentally demonstrate that OPR can correct up to 72% of packets that would normally have failed, when they are captured by multiple gateways.

A comparative study of location-sharing privacy preferences in the United States and China

Lin J., Benisch M., Sadeh N., Niu J., Hong J., Lu B., Guo S.

Knowledge and Information Systems volume

2016

Abstract:

While prior studies have provided us with an initial understanding of people’s location-sharing privacy preferences, they have been limited to Western countries and have not investigated the impact of the granularity of location disclosures on people’s privacy preferences. We report findings of a 3-week comparative study collecting location traces and location-sharing preferences from two comparable groups in the United States and China. Results of the study shed further light on the complexity of people’s location-sharing privacy preferences and key attributes influencing willingness to disclose locations to others and to advertisers. While our findings reveal many similarities between US and Chinese participants, they also show interesting differences, such as differences in willingness to share location at “home” and at “work” and differences in the granularity of disclosures people feel comfortable with. We conclude with a discussion of implications for the design of location-sharing applications and location-based advertising.

A computationally efficient implementation of fictitious play in a distributed setting

Swenson B., Kar S., Xavier J.

2015 23rd European Signal Processing Conference, EUSIPCO 2015

2015

Abstract:

The paper deals with distributed learning of Nash equilibria in games with a large number of players. The classical fictitious play (FP) algorithm is impractical in large games due to demanding communication requirements and high computational complexity. A variant of FP is presented that aims to mitigate both issues. Complexity is mitigated by use of a computationally efficient Monte-Carlo based best response rule. Demanding communication problems are mitigated by implementing the algorithm in a network-based distributed setting, in which player-to-player communication is restricted to local subsets of neighboring players as determined by a (possibly sparse, but connected) preassigned communication graph. Results are demonstrated via a simulation example.

A fast spatial variation modeling algorithm for efficient test cost reduction of analog/RF circuits

Goncalves H., Li X., Correia M., Tavares V., Carulli J., Butler K.

Proceedings -Design, Automation and Test in Europe, DATE

2015

Abstract:

In this paper, we adopt a novel numerical algorithm, referred to as dual augmented Lagrangian method (DALM), for efficient test cost reduction based on spatial variation modeling. The key idea of DALM is to derive the dual formulation of the L1-regularized least-squares problem posed by Virtual Probe (VP), which can be efficiently solved with substantially lower computational cost than its primal formulation. In addition, a number of unique properties associated with discrete cosine transform (DCT) are exploited to further reduce the computational cost of DALM. Our experimental results of an industrial RF transceiver demonstrate that the proposed DALM solver achieves up to 38x runtime speed-up over the conventional interior-point solver without sacrificing any performance on escape rate and yield loss for test applications.

A framework for actuator placement in large scale power systems: Minimal strong structural controllability

Pequito S., Popli N., Kar S., Ilic M.D., Aguiar A.P.

2013 5th IEEE International Workshop on Computational Advances in Multi-Sensor Adaptive Processing, CAMSAP 2013

2013

Abstract:

This paper addresses the problem of minimal placement of actuators in large-scale linear time invariant (LTI) systems, such as large-scale power systems, for dynamic controller design. A novel sufficient and necessary condition to ensure a strong structurally controllable (SSC) system is proposed. Specifically, the paper addresses the problem of obtaining the minimal number of dedicated inputs, i.e., inputs which actuate only a single state variable, and the respective state variables they should be assigned to, such that the LTI system is SSC. In addition, an efficient and scalable algorithm, with polynomial implementation complexity, to achieve such minimal placement of dedicated inputs is proposed. An illustration of the proposed design methodology is provided on the IEEE 5-bus test system, thereby identifying the minimal number of physical state variables to be actuated for ensuring strong structural controllability.