Search

Controversial or complex topics often exhibit the backfire effect, where users' opinions harden in the face of facts to the contrary. We present initial work towards developing an online social collaborative argumentation system to verify alternative facts and misinformation by also including users' emotional associations with those stances. Our goal is to help users more effectively explore and understand their possibly subconscious biases in an effort to overcome the backfire effect and formulate more varied insights into complex and controversial topics. In order to aid this process, we model their emotional profile on such topics and combine it with a proposition profile, based on the semantic and collaborative content of propositions. We develop an algorithm to generate sentiment-based models of claims and propositions which we can filter based on users' inferred beliefs and the strength of those beliefs.

We present two artistically-relevant algorithms to aid in the quantification of artistic style, the Discrete Tonal Measure (DTM) and Discrete Variational Measure (DVM). These quantitative features can provide clues to the artistic elements that enable art scholars to categorize works as belonging to different artistic styles. We also introduce two new datasets for analysis of artistic style: one based on the school of art to which artists belong and one based on the medium used by a specific artist. We show results of initial experiments for classifying paintings on each of these datasets with DTM and DVM using a scientific workflows framework that will allow reuse and extension of many visual stylometry methods, as well as allowing easy reproducibility of analytical results, by publishing datasets and workflows packaged as linked data.

Reproducibility of research is an area of growing concern in computer vision. Scientific workflows provide a structured methodology for standardized replication and testing of state-of-the-art models, open publication of datasets and software together, and ease of analysis by re-using pre-existing components. In this paper, we present initial work in developing a framework that will allow reuse and extension of many computer vision methods, as well as allowing easy reproducibility of analytical results, by publishing dadasets and workflows packaged together as linked data. Our approach uses the WINGS semantic workflow system which validates semantic constraints of the computer vision algorithms, making it easy for non-experts to correctly apply state-of-the-art image processing methods to their data. We show the ease of use of semantic workflows for reproducibility in computer vision by both utilizing pre-developed workflow fragments and developing novel computer vision workflow fragments for a video activity recognition task, analysis of multimedia web content, and the analysis of artistic style in paintings using convolutional neural networks.

Collaboration, extension, and reproduction of research is of great importance in computer vision. Scientific workflows offer a unique framework for distributed collaboration and sharing of experiments. They provide a structured, end-to-end analysis methodology that easily and automatically allows for standardized replication and testing of models, inter-operability of heterogeneous codebases, and incorporation of novel algorithms. In this paper, we introduce the use of scientific workflows in computer vision to aid collaboration.

Understanding group activities is an essential step towards studying complex crowd behaviours in video. However, such research is often hampered by the lack of a formal definition of a group, as well as a dearth of datasets that concentrate specifically on Atomic Group Actions. 1 In this paper, we provide a quantitative definition of a group based on the Group Transition Ratio (G tr ); the G tr helps determine when individuals transition to becoming a group (where the individuals can still be tracked) or a crowd (where tracking of individuals is lost). In addition, we introduce the Atomic Group Actions Dataset, a set of 200 videos that concentrate on the atomic group actions of objects in video, namely the group-group actions of formation, dispersal, and movement of a group, as well as the group-person actions of person joining and person leaving a group. We further incorporate a structured, end-to-end analysis methodology, based on workflows, to easily and automatically allow for standardized testing of new group action models against this dataset. We demonstrate the efficacy of the G tr on the Atomic Group Actions Dataset and make the full dataset (the videos, along with their associated tracks and ground truth, and the exported workflows) publicly available to the research community for free use and extension at at http://research. sethi.org/ricky/datasets/.

Semantic wikis augment wikis with semantic properties that can be used to aggregate and query data through reasoning. Semantic wikis are used by many communities, for widely varying purposes such as organizing genomic knowledge, coding software, and tracking environmental data. Although wikis have been analyzed extensively, there has been no published analysis of the use of semantic wikis. In this paper, we analyze twenty semantic wikis selected for their diverse characteristics and content. We analyze the property edits and compare to the total number of edits in the wiki. We also show how semantic properties are created over the lifetime of the wiki.

Motion and image analysis are both important for robust solutions to video search of activities; the physics-based, data-driven Hamiltonian Monte Carlo (HMC), a Markov chain Monte Carlo variant that is efficient in searching large dimensional spaces, simultaneously examines the combined motion and image space. In this paper, we generalize the data-driven HMC to no longer depend upon ad hoc Guide Hamiltonians and to no longer require physics-based features from tracks as pre-requisites. Our generalization thus allows it to be used with or without a tracker, overcoming a significant limitation of the physics-based approach, as well as being extensible to utilizing any pre-existing image- or motion-based method. We demonstrate the generalizability of our framework by considering situations when tracking is available and when it is not available. When tracking is available, we utilize Histogram of Oriented Gradients, shapes of trajectories, and Hamiltonian Energy Signatures; when tracking is not available, we use Space-time Interest Points and GIST features. In addition, we show our generalized framework performs better than the physics-based, data-driven HMC, as well as state-of-the-art, by demonstrating the efficacy of our system on real-life video sequences using the well-known Weizmann and YouTube Action datasets.

In this paper, we demonstrate the ability to reuse workflow fragments in different data domains: from text analytics to image analysis to video activity recognition. We highlight how the re-use of workflows allows scientists to link across disciplines and avail themselves of the benefits of interdisciplinary research beyond their normal area of expertise. In addition, we present an in-depth study of a Social Media Analysis (SMA) task, wherein we show how the re-use of workflow fragments can extend a pre-existing, rudimentary analysis; we also examine how workflow fragments save time and effort in SMA while bringing together multiple areas of machine learning and computer vision.

This paper concentrates on the problem of modelling and recognition of complex behaviours involving multi-object interactions in video. We use motion patterns of individual objects to construct models which characterize pairs by correlating them in phase space. These models of complex interactions allow for: recognition of group activities, which occur when individual people or objects start interacting as a single entity; detection of transitions from individuals to groups to crowds; and the interactions of individuals with groups, as well as the interactions of groups with other groups. We establish a general formalism by examining activities using relative distances and analyse multi-object interactions directly via the Phase Space Algorithm. Finally, we calculate a scale-invariant Group Transition Ratio to quantify formation and dispersal of both groups and crowds. Our input is solely the position information of individuals, which we get using a person tracker, optical flow, and Lagrangian particle dynamics. We demonstrate the uses of this model for recognition of complex activities on the standard CAVIAR, VIVID, and UCR Videoweb datasets.

In this paper, we develop a neurobiologically-motivated statistical method for video analysis that simultaneously searches the combined motion and form space in a concerted and efficient manner using well-known Markov chain Monte Carlo (MCMC) techniques. Specifically, we leverage upon an MCMC variant called the Hamiltonian Monte Carlo (HMC), which we extend to utilize data-based proposals rather than the blind proposals in a traditional HMC, thus creating the Data-Driven HMC (DDHMC). We demonstrate the efficacy of our system on real-life video sequences.

Recognizing a person's motion is intuitive for humans but represents a challenging problem in machine vision. In this paper, we present a multi-disciplinary framework for recognizing human actions. We develop a novel descriptor, the Human Action Image (HAI): a physically-significant, compact representation for the motion of a person, which we derive from first principles in physics using Hamilton's Action. We embed the HAI as the Motion Energy Pathway of the latest Neurobiological model of motion recognition. The Form Pathway is modelled using existing low-level feature descriptors based on shape and appearance. Experimental validation of the theory is provided on the well-known Weizmann and USF Gait datasets.

In this paper, we propose a computational framework for integrating the physics of motion with the neurobiological basis of perception in order to model and recognize human actions and object activities. The essence, or gist, of an action is intrinsically related to the motion of the scene's objects. We define the Hamiltonian Energy Signature (HES) and derive the S-Metric to yield a global representation of the motion of the scene's objects in order to capture the gist of the activity. The HES is a scalar time-series that represents the motion of an object over the course of an activity and the S-Metric is a distance metric which characterizes the global motion of the object, or the entire scene, with a single, scalar value. The neurobiological aspect of activity recognition is handled by casting our analysis within a framework inspired by Neuromorphic Computing (NMC), in which we integrate a Motion Energy model with a Form/Shape model. We employ different Form/Shape representations depending on the video resolution but use our HES and S-Metric for the Motion Energy approach in either case. As the core of our Integration mechanism, we utilize variants of the latest neurobiological models of feature integration and biased competition, which we implement within a Multiple Hypothesis Testing (MHT) framework. Experimental validation of the theory is provided on standard datasets capturing a variety of problem settings: single agent actions (KTH), multi-agent actions, and aerial sequences (VIVID).