Computer Science and Engineering

<p>Shiv Nadar University is a comprehensive, multidisciplinary, research-focused, and student-centric University offering a full range of academic programs at the undergraduate, postgraduate and doctoral level. With state-of-the-art infrastructure, the University comprises of academic wings, sophisticated labs, international standard sports facilities, amphitheaters, auditorium, conference rooms and smart classrooms. The University&rsquo;s goal is to become internationally recognized for the quality of its research and creative endeavors and their applicability to improving quality of life, generating new insights and expanding the boundaries of human knowledge creativity. Committed to excellence in teaching, research and service, the University aims to serve the higher education needs of India and the world beyond.</p>

Shiv Nadar University

Face recognition in images is an active area of interest among the computer vision researchers. However, recognizing human face in an unconstrained environment, is a relatively less-explored area of research. Multiple face recognition in unconstrained environment is a challenging task, due to the variation of view-point, scale, pose, illumination and expression of the face images. Partial occlusion of faces makes the recognition task even more challenging. The contribution of this paper is two-folds: Introducing a challenging multi-face dataset (i.e., IIITS-MFace Dataset) for face recognition in unconstrained environment and evaluating the performance of state-of-the-art hand-designed and deep learning based face descriptors on the dataset. The proposed IIITS-MFace dataset contains faces with challenges like pose variation, occlusion, mask, spectacle, expressions, change of illumination, etc. We experiment with several state-of-the-art face descriptors, including recent deep learning based face descriptors like VGGFace, and compare with the existing benchmark face datasets. Results of the experiments clearly show that the difficulty level of the proposed dataset is much higher compared to the benchmark datasets. © 2018 IEEE.

World University root node

(SoE) School of Engineering

2018 15th International Conference on Control, Automation, Robotics and Vision, ICARCV 2018

A Multi-Face Challenging Dataset for Robust Face Recognition

The attractiveness of a baby face image depends on the perception of the perceiver. However, several recent studies advocate the idea that human perceptual analysis can be approximated by statistical models. We believe that the cuteness of baby faces depends on the low level facial features extracted from different parts (e.g., mouth, eyes, nose) of the faces. In this paper, we introduce a new problem of classifying baby face images based on their cuteness level using supervised learning techniques. The proposed learning model finds the potential of a deep learning technique in measuring the level of cuteness of baby faces. Since no datasets are available to validate the proposed technique, we construct a dataset of images of baby faces, downloaded from the internet. The dataset consists of several challenges like different view-point, orientation, lighting condition, contrast and background. We annotate the data using some well-known statistical tools inherited from Reliability theory. The experiments are conducted with some well-known image features like Speeded Up Robust Feature (SURF), Histogram of Oriented Gradient (HOG), Convolutional Neural Network (CNN) on Gradient and CNN on Laplacian, and the results are presented and discussed. © 2017, Springer Science+Business Media, LLC.

Multimedia Tools and Applications

Measuring level of cuteness of baby images: a supervised learning scheme

We propose a unified method for recognizing human action and human related events in a realistic video. We use an efficient pipeline of (a) a 3D representation of the Improved Dense Trajectory Feature (DTF) and (b) Fisher Vector (FV). Further, a novel descriptor is proposed, capable of representing human actions and human related events based on the FV representation of the input video. The proposed unified descriptor is a 168-dimensional vector obtained from each video sequence by statistically analyzing the motion patterns of the 3D joint locations of the human body. The proposed descriptor is trained using binary Support Vector Machine (SVM) for recognizing human actions or human related events. We evaluate the proposed approach on two challenging action recognition datasets: UCF sports and CMU Mocap datasets. In addition to the two action recognition dataset, the proposed approach is tested on the Hollywood2 event recognition dataset. On all the benchmark datasets for both action and event recognition, the proposed approach has shown its efficacy compared to the state-of-the-art techniques. © 2017, Springer Science+Business Media, LLC.

Human action and event recognition using a novel descriptor based on improved dense trajectories

Analyzing the lifecycle of topics, that are present in user-generated text content, has emerged as a mainstream topic of social network research. The literature presently identifies topics on Twitter, a prominent online social network, as either individual hashtags, or a burst of keywords within a short span of time, or as latent concept spaces obtained from sophisticated text analysis mechanisms, such as Latent Dirichlet Allocation (LDA). The first and second approaches fail to recognize that topics do not restrict themselves to individual hashtags and are likely to span across (semantically related) keywords, while the third does not capture the user’s intended topics expressed via hashtags. In the current paper, we propose a novel methodology that addresses these shortcomings. We jointly utilize the temporal concurrency of the hashtags contained in given tweets and the latent concept space addressed by the tweet content, to identify groups of hashtags representing concept space—a “topic”—addressed by many tweets. A given topic, thus, is represented by a different set of representative hashtags at different times; the usage rate of the different hashtags change such that some hashtags gain prominence over others over time. Unlike the literature, where lifecycle analysis of one topic typically comprises of analyzing one hashtag, we analyze and characterize the lifecycle of a topic as a combination of multiple semantically and temporally related hashtags. We derive novel insights about lifecyle of topics: the inception and continuity of the topics over time (expressed over different hashtags), and how topics morph over hashtags, from one set of hashtags to another, before eventually dying down. © Springer International Publishing AG 2018.

Studies in Computational Intelligence

A semantic continuity based analysis of topic lifecycle on social networks

Background: Topic lifecycle analysis on social networks aims to analyze and track how topics are born from user-generated content, and how they evolve. Twitter researchers have no agreed-upon definition of topics; topics on Twitter are typically derived in the form of (a) frequently used hashtags, or (b) keywords showing sudden trends of large occurrence in a short span of time (“bursty keywords”), or (c) concepts latent within the tweets that are grouped using variations of semantic clustering techniques. Methods: In the current paper, we jointly model the hashtags present and the semantic concepts embedded in the content, which in turn helps us identify hashtag groups that define a “topic”—a concept space—that are used by a large number of tweets. Results: We observe that different hashtags belonging to a given cluster are more prominent compared to the others, at different times. We further observe that the participation and influence levels of the different users play important roles in determining which hashtag would be more prominent than the others at given times. We thus observe topics to often morph from one to the other (via morphing of dominant hashtags representing the same semantic concept space), rather than becoming extinct outright, which is a novel insight about topic lifecycles. We further present novel observations about the role of users in determining the lifecycle of discussion topics on Twitter. Conclusions: We infer that topic lifecycles are governed by user interests, and not by user influence, which is a key observation made by our work. © 2018, The Author(s).

Fulltext

Computational Social Networks

Assessing the role of participants in evolution of topic lifecycles on social networks

Homophily, the phenomenon of similar people getting connected to and being socially familiar with each other, is well-known on online social networks. Detection of user stance towards given topics, on online social networks, specifically Twitter, has emerged as a mainstream research topic. The current work provides insights into the impact of topic-specific stance similarity and social familiarity on social interaction dynamics. This is a novel and yet fundamental problem in social networks research, that has so far remained unexplored in the literature. Specifically, we address two key aspects. One, we investigate whether the smoothness (politeness) level of conversations between user pairs, relate with overall stance similarity (spanning across topics). Two, we examine the impact on interaction smoothness (politeness) with respect to social familiarity and topical stance-similarity. We propose a novel approach based on word embedding, to compare across users and across topics. We analyze the relationship between topical stance similarity, social familiarity and interaction politeness of users, with respect to specific familiarities between user pairs as well as social communities. © Springer International Publishing AG 2018.

Assessing the effects of social familiarity and stance similarity in interaction dynamics

In densely populated metropolitan cities, people spend an average of 45-150 minutes inside a vehicle commuting to their workplace. Most of these travelers often experiences short term health conditions like drowsiness, breathlessness, discomfort and headache. Many accidents have been reported due to drowsiness while driving. There has been considerable increase in the number of death/choking cases reported due to CO 2 buildup inside the vehicle. Even though Vehicle Indoor Air Quality is an important parameter, the general awareness about it is very low, and only few high end new models of automobiles have indoor air quality sensing systems, with limited capability. This paper attempts to use a sensing system that can be retrofitted in any old or new vehicle to measure the CO 2 level inside the vehicle and use this sensed data to perform statistical and stochastic analysis of the growth pattern of CO 2 inside the vehicle chamber. © 2018 IEEE.

Proceedings of IEEE Sensors

Predictive Modelling of In-Vehicle CO 2 Concentration Using Sensor Data Analytics

Scheduling a meeting is a difficult task for people who have overbooked calendars and often have many constraints. This activity becomes further complex when the meeting is to be scheduled between parties who are situated in geographically distant locations of a city and have varying traveling patterns. To achieve this, we first propose a solution to determine optimal meeting location for two moving users in the Euclidean space. Then, we generalize the problem by considering variable number of moving users and evaluate optimal meeting point on the road network. We extend the work of Yan et al. (Proc VLDB Endow 4(11):1–11, 2011) in this domain by incorporating some real life constraints like variable number of users, varying travel patterns, flexible meeting point and considering road network distance. Experiments are performed on a real-world dataset and show that our method is effective in stated conditions. © 2018, Springer-Verlag GmbH Austria, part of Springer Nature.

Social Network Analysis and Mining

Mining optimal meeting points for moving users in spatio-temporal space

Assessment of the Translation done by machines has always been a major topic of interest for the Natural Language Researchers. Every day, a new translator is out there in the market claiming to be the best one. But to challenge such claims, one needs to have a powerful assessment tool that can judge the output of any translator on different parameters like fluency of the language, adequacy as well as the accuracy of the output generated. In this paper, we have discussed the major issues with the algorithms used for the evaluation of Machine translation. We have also presented a brief description and comparative study of different metrics used to evaluate the outcome of a machine translator over English to Hindi Language parallel sentences. © 2018 IEEE.

Proceedings of the 8th International Conference Confluence 2018 on Cloud Computing, Data Science and Engineering, Confluence 2018

A Summary and Comparative Study of Different Metrics for Machine Translation Evaluation

IoT devices generate data at a remarkable speed which requires near real-time processing. Such need has inspired a new computing paradigm that advocates moving computation to the edge, closer to where data is generated for ensuring low-latency and responsive data analytics. This paper aims at monitoring indoor environment parameters inside the classroom in real time and estimate the occupancy of that room based on the collected data This is done by measuring different environmental parameters like temperature, CO2 etc. and then by performing regression on the collected data. For the purpose of real time monitoring an android application is used which uses MQTT for communication with the edge node. The estimation is then verified against the actual attendance in the classroom at that duration and it gives accuracy between 90\%-95\%. © 2018 IEEE.