Department of Computer Science & Engineering

&ldquo;IIT Jodhpur is committed to impart quality education to facilitate the creation of qualified, competent and responsible human resources to meet the emerging technological challenges of the world. It is working towards creating an equitable learning platform that is positive, accessible and effective in the face of the modern societal and scientific advancements.

Flexibility in curriculum, in all its academic programs, encompassing elements of teaching, research, creativity, innovation and entrepreneurship is the most unique feature of IIT Jodhpur. The Department of Science and Technology (DST), under the National Mission on Cyber Physical Systems, has identified IIT Jodhpur as a Technology &amp; Innovation Hub in the area of Augmented Reality and Virtual Reality. The institute has also been nominated as the coordinator for the Jodhpur City and Knowledge Cluster by the Office of Principal Scientific Adviser, Government of India.

The sprawling 852-acre IIT Jodhpur campus is a majestic tribute to the deep-rooted diverse cultural heritage of the region. The campus is an architectural masterpiece, which is slated to become an international exemplar of sustainability with net-zero energy, and advanced water and waste management strategies.&rdquo;

IIT Jodhpur

Searching is one of the fundamental tasks in Computer Science. An intuitive way to search is to do it linearly, that is, start at the beginning of the dataset and continue till the searched-for item is found or nothing is found. However, as the volume of data increases, the response time of linear search is no longer acceptable. Indexes are designed to search through massive datasets quickly. There are a number of different ways of building complex and advanced indexes. Appropriate selection and modification of indexing structures according to dynamic business requirements is crucial for data-intensive applications. In this work, we present a few basic reusable indexing structures. These structures can be used to create advanced and complex indexing structures with lesser effort and time. © 2018 Copyright held by the owner/author(s). Publication rights licensed to ACM.

World University root node

ACM International Conference Proceeding Series

Patterns for indexing large datasets

We study the problem of retrieving cartoon faces of celebrities given their real face as a query. We refer to this problem as Photo2Cartoon. The Photo2Cartoon problem is challenging since (i) cartoons vary excessively in style and (ii) modality gap between real and cartoon faces is large. To address these challenges, we present a discriminative deep metric learning approach designed for matching cross-modal faces and showcase Photo2Cartoon. The proposed approach learns a nonlinear transformation to project real and cartoon face pairs into a common subspace where distance between positive pairs becomes smaller as compared to distance between negative pairs. We evaluate our method on two public benchmarks, namely IIIT-CFW and Viewed Sketch, and show superior retrieval results as compared to related methods. © 2018, Springer-Verlag London Ltd., part of Springer Nature.

International Journal of Multimedia Information Retrieval

DHFML: deep heterogeneous feature metric learning for matching photograph and cartoon pairs

This paper describes a new approach of viewing a social relation as a string with various forces acting on it. Accordingly, a tension measure for a relation is defined. Various component forces of the tension measure are identified based on the structural information of the network. A new variant of rough set, namely, double bounded rough set, is developed in order to define these forces mathematically. It is revealed experimentally with synthetic and real-world data that positive and negative tension characterizes, relatively, the presence and absence of a physical link between two nodes. An algorithm based on tension measure is proposed for link prediction. Superiority of the algorithm is demonstrated on nine real-world networks, which include four temporal networks. The source code for calculating tension measure and link prediction algorithm is publicly available at https://gitlab.com/suman5/social-tension-measure. © 2014 IEEE.

IEEE Transactions on Computational Social Systems

Double bounded rough set, tension measure, and social link prediction

Customer segmentation based on temporal variation of subscriber preferences is useful for communication service providers (CSPs) in applications such as targeted campaign design, churn prediction, and fraud detection. Traditional clustering algorithms are inadequate in this context, as a multidimensional feature vector represents a subscriber profile at an instant of time, and grouping of subscribers needs to consider variation of subscriber preferences across time. Clustering in this case usually requires complex multivariate time series analysis-based models. Because conventional time series clustering models have limitations around scalability and ability to accurately represent temporal behaviour sequences (TBS) of users, that may be short, noisy, and non-stationary, we propose a latent Dirichlet allocation (LDA) based model to represent temporal behaviour of mobile subscribers as compact and interpretable profiles. Our model makes use of the structural regularity within the observable data corresponding to a large number of user profiles and relaxes the strict temporal ordering of user preferences in TBS clustering. We use mean-shift clustering to segment subscribers based on their discovered profiles. Further, we mine segment-specific association rules from the discovered TBS clusters, to aid marketers in designing intelligent campaigns that match segment preferences. Our experiments on real world data collected from a popular Asian communication service provider gave encouraging results. Copyright © 2017 John Wiley & Sons, Ltd

Expert Systems

Clustering short temporal behaviour sequences for customer segmentation using LDA

Stomatal morphology is a key phenotypic trait for plants' response analysis under various environmental stresses (e.g. Drought, salinity etc.). Stomata exhibit diverse characteristics with respect to orientation, size, shape and varying degree of papillae occlusion. Thus, the biologists currently rely on manual or semi-automatic approaches to accurately compute its morphological traits based on scanning electron microscopic (SEM) images of leaf surface. In contrast to these subjective and low-throughput methods, we propose a novel automated framework for stomata quantification. It is realized based on a hybrid approach where the candidate stomata region is first detected by a convolutional neural network (CNN) and the occlusion is dealt with an inpainting algorithm. In addition, we propose stomata segmentation based quantification framework to solve the problem of shape, scale and occlusion in an end-to-end manner. The performance of the proposed automated frameworks is evaluated by comparing the derived traits with manually computed morphological traits of stomata. With no prior information about its size and location, the hybrid and end-to-end machine learning frameworks shows a correlation of 0.94 and 0.93, respectively on rice stomata images. Furthermore, they successfully enable wheat stomata quantification showing generalizability in terms of cultivars. © 2018 IEEE.

Proceedings - International Conference on Pattern Recognition

Automatic Quantification of Stomata for High-Throughput Plant Phenotyping

Compressed Sensing (CS) has emerged as an alternate method to acquire high dimensional signals effectively by exploiting the sparsity assumption. However, owing to non-sparse and non-stationary nature, it is extremely difficult to process Electroencephalograph (EEG) signals using CS paradigm. The success of Bayesian algorithms in recovering non-sparse signals has triggered the research in CS based models for neurophysiological signal processing. In this paper, we address the problem of Temporal Modeling of EEG Signals using Block Sparse Variational Bayes (SVB) Framework. Temporal correlation of EEG signals is modeled blockwise using normal variance scale mixtures parameterized via some random and deterministic parameters. Variational inference is exploited to infer the random parameters and Expectation Maximization (EM) is used to obtain the estimate of deterministic parameters. To validate the framework, we present experimental results for benchmark State Visual Evoked Potential (SSVEP) dataset with 40-target Brain-Computer Interface (BCI) speller using two frequency recognition algorithms viz. Canonical Correlation Analysis (CCA) and L1-regularized Multiway CCA. Results show that the proposed temporal model is highly useful in processing SSVEP-EEG signals irrespective of the recognition algorithms used. © 2018 ACM.

Temporal Modeling of EEG Signals using Block Sparse Variational Bayes Framework

In this letter, we aim to develop a deep adversarial despeckling approach to enhance the quality of ultrasound images. Most of the existing approaches target a complete removal of speckle, which produces oversmooth outputs and results in loss of structural details. In contrast, the proposed approach reduces the speckle extent without altering the structural and qualitative attributes of the ultrasound images. A despeckling residual neural network (DRNN) is trained with an adversarial loss imposed by a discriminator. The discriminator tries to differentiate between the despeckled images generated by the DRNN and the set of high-quality images. Further to prevent the developed network from oversmoothing, a structural loss term is used along with the adversarial loss. Experimental evaluations show that the proposed DRNN outperforms the state-of-the-art despeckling approaches in terms of the structural similarity index measure, peak signal to noise ratio, edge preservation index, and speckle region's signal to noise ratio. © 1994-2012 IEEE.

IEEE Signal Processing Letters

Ultrasound image enhancement using structure oriented adversarial network

The important aspect of intangible heritage are the classical stories and folklores which have been passed down over the ages through classical texts, scriptures and word of mouth. The story of Girija Kalyana, which describes various events linked to the marriage of Indian goddess Parvati to God Shiva, is one of such stories from the intangible heritage of India. Every such story has numerous manifestations in tangible heritage artefacts such as statues depicting story characters; mural paintings depicting events in the story; monuments and locations related to aspects of the storyline. Similar depictions of the story are found in intangibles like dance performances, theatre, folk-art and living traditions. In this chapter, we discuss a new paradigm in digital heritage preservation which engages semantic web techniques to offer an immersive and experiential exploration of a heritage theme. A story can have different narratives, each of which, offers a different perspective with reflections in mythology, art and architecture, literature, rituals and crafts, and social and geographical linkages. These narratives can be presented to users with the help of a reasoning framework, based on a domain ontology which provides the means to encode the rich, scholarly knowledge existing with cultural heritage experts. The ontology also enables semantic linkages between digital artefacts in various media formats through curated, semantic annotation. Multimedia Web Ontology Language (MOWL) representation of the heritage ontology permits perceptual modelling of domain concepts in terms of their observable media properties. The language has media propagation rules which allow specification of a semantic concept to be extended to its related concepts in the ontology, and thus helps to widen the scope of concept recognition in media. The ontology-based framework provides a digital infrastructure for a virtual user interface offering an intellectual exploration of the heritage theme in space and time with ontology-guided navigation. © Springer Nature Singapore Pte Ltd. 2017.

Digital Hampi: Preserving Indian Cultural Heritage

Ontology-based narratives of the Girija Kalyana

Ultrasound (US) image despeckling is a problem of high clinical importance. Machine learning solutions to the problem are considered impractical due to the unavailability of speckle-free US image dataset. On the other hand, the classical approaches, which are able to provide the desired outputs, have limitations like input dependent parameter tuning. In this work, a convolutional neural network (CNN) is developed which learns to remove speckle from US images using the outputs of these classical approaches. It is observed that the existing approaches can be combined in a complementary manner to generate an output better than their individual outputs. Thus, the CNN is trained using the individual outputs as well as the output ensembles. It eliminates the cumbersome process of parameter tuning required by the existing approaches for every new input. Further, the proposed CNN is able to outperform the state-of-the-art despeckling approaches and produces the outputs even better than the ensembles for certain images. © 2018 IEEE.

Despeckling CNN with Ensembles of Classical Outputs

Continuous efforts are being made to understand human perception network with the purpose of developing enhanced computational models for vision-based tasks. In this paper, we utilize eye gaze as a medium to unravel the cues utilized by humans for the perception of facial aging. Specifically, we explore the tasks of face age estimation and age-separate face verification and analyze the eye gaze patterns of participants to understand the strategy followed by human participants. To facilitate this, eye gaze data from 50 participants is acquired using two different eye gaze trackers: Eye Tribe and GazePoint GP3. Comprehensive analysis of various eye movement metrics is performed with respect to different face parts to illustrate their relevance for age estimation and age-separated face verification tasks. © 2018 IEEE.