Computer Science and Engineering

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.

Shiv Nadar University

Selecting a business site is a significant decision in managing a commercial organization. It becomes more complex in countries like India, where population density is high, and it is tough to distribute services fairly. With this difficulty, investing in a location without a thorough survey is challenging. This study aims to provide an information map that can help choose the optimal neighborhood for a particular business. The research aims to use data science powers to provide an unsupervised machine-learning approach to offer a recommendation-based solution that suggests the best location for a given use case. The objective of this paper is the in-depth analysis for site-finding that includes clustering regions based on various factors such as geographical locations (like schools, bus and railway stations, and hospitals), sanitation, distance, footfall, and other entities. The idea is to obtain the user's use case query and then create hotspots on the map that show the most suitable locations where the specific use case will benefit the stakeholders the most. We need to identify neighborhoods where there may be an unsatisfied need for Indian restaurants. Additionally, we require places that are uncrowded and have little competition. If the first two requirements are satisfied, we would also like a location close to a prominent city neighborhood. The advantages of each area will then be clearly expressed so that stakeholders can choose the best possible final location.

(SoE) School of Engineering

2024 IEEE International Conference on Contemporary Computing and Communications

Location-Based Ideal Site Selection using Clustering

This research explores the intricate relationship between weather-related variables and the transmission dynamics of two pivotal diseases: COVID-19 and dengue. Our primary goal is to advance our understanding of how climate factors influence the occurrence and spread of these diseases through a comparative analysis. In the context of the pressing need for accurate prediction systems, our research aims to contribute to the refinement of public health planning and decision-making strategies. In this work, our findings underscore the superiority of certain models in predicting epidemic outbreaks. Notably, Random Forest consistently demonstrated exceptional performance across all metrics for both COVID-19 and Dengue predictions. The model exhibited superior accuracy, precision, and correlation with actual values.

Asian Conference on Intelligent Information and Database Systems

Predicting Epidemic Outbreak Using Climatic Factors

Diffusion tensor imaging (DTI) is a non-invasive technique for analyzing the movement of water in the brain. However, the precision of measurements required for tracking white matter pathways can lead to long scan times, which can be challenging for some patient populations such as pediatric patients. To address this issue, researchers have been experimenting with deep learning techniques for faster estimation of DTI parameters, which are helpful in neurological diagnosis, of diffusion-weighted images. Our proposed solution is a transformer neural network-based approach for fast estimation of diffusion tensor parameters using sparse measurements. While there have been attempts to address this problem, our proposed model handles both scalable and generalized estimation of DTI parameters using multiple sparse measurements. Through experimentation on the Human Connectome Project (HCP) Young Adult benchmark dataset, our proposed model demonstrated state-of-the-art results in terms of fractional anisotropy (FA), axial diffusivity (AD), and mean diffusivity (MD) when compared to traditional linear least square (LLS) fitting and 3D U-Net model with input size (3D U-Net16).

Neural Computing and Applications

SwinDTI: Swin transformer based generalized fast estimation of diffusion tensor parameters from sparse data

Alzheimer disease is one of the most common neuro-degenerative diseases, with an estimated 6.2 million cases in the United States. This research article investigates the potential of Transformer-based deep learning techniques to accelerate the processing of diffusion tensor imaging (DTI) measures and improve the early diagnosis of Alzheimer disease (AD) using sparse data. Diffusion Weighted Imaging (DWI) is a time-consuming process, with each diffusion direction taking between 2-5 minutes, and at least 40 diffusion directions are needed for routine clinical diagnosis, which needs scanning duration exceeding 3 hours for each patient. By leveraging the attention mechanism, our proposed model generates quantitative measures of fractional anisotropy (FA), axial diffusivity (AxD), and mean diffusivity (MD) using 5 and 21 diffusion directions, making it useful for clinical diagnosis through reduced scanning time of more than half. Our experimental results on the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset demonstrate that our proposed model outperforms the traditional linear least square method, achieving accurate quantitative measurement of FA, AxD, and MD scores for early diagnosis of AD patients from healthy controls using sparse diffusion directions. Our analysis highlights the potential of Swin-Transformer attention-based deep learning framework to improve the early diagnosis and treatment of Alzheimer’s disease.

Proceedings of Machine Learning Research

Early Diagnosis of Alzheimer through Swin-Transformer-Based Deep Learning Framework using Sparse Diffusion Measures

Frontotemporal dementia (FTD) is a devastating neurodegenerative disorder that primarily affects the frontal and temporal lobes of the brain, leading to cognitive decline and behavioral changes. Early and accurate diagnosis of FTD is crucial for initiating timely interventions and providing appropriate care to patients. In the opinion of the experts, about 12-22 persons out of the population of 100,000 persons experience FTD. That means between 1.2 million and 1.8 million people have it worldwide. This research paper proposes a novel deep learning framework that utilizes sparse diffusion measures extracted from neuroimaging data to aid in the early diagnosis of Frontotemporal Dementia. The proposed model leverages the power of deep learning techniques to automatically learn relevant features from the data and effectively distinguish between healthy individuals and those with FTD. The experimental results demonstrate the promising potential of the proposed approach in improving FTD diagnosis and paving the way for future research in this area.

IEEE/CVF International Conference on Computer Vision

Deep Learning Framework using Sparse Diffusion MRI for Diagnosis of Frontotemporal Dementia

This paper aims to improve the identification of taxi hotspots in spatio-temporal space. We propose an approach that combines ensemble-weighted degree and node entropy measures within a dominating set to enhance the identification of taxi hotspots. First, we construct a graph representation of the spatio-temporal space, where nodes represent potential taxi hotspots. The weighted degree of each node is computed by considering the weights of its adjacent edges, which correspond to the distances between the two adjacent nodes on the road network. This measure quantifies the importance and connectivity of a node. We calculate node entropy to capture the level of uncertainty and randomness associated with each node. The entropy measure provides insights into the diversity of each potential hotspot, helping to distinguish between highly active and less predictable areas. We then integrate the ensemble weighted degree and node entropy measures into a dominating set approach. By iteratively selecting the nodes with the highest combined scores, we construct an optimal set of taxi hotspots that effectively cover the spatio-temporal space while considering both connectivity and uncertainty. The proposed approach enables a more comprehensive identification of taxi hotspots, taking into account the connectivity and importance of potential locations and their level of uncertainty. To evaluate the effectiveness of the proposed approach, we conducted experiments using New York taxi trip data. The results demonstrate that the ensemble weighted degree and node entropy measures enhance the identification of taxi hot spots compared to traditional dominating set methods.

<a href="https://link.springer.com/chapter/10.1007/978-3-031-48316-5_31">Enhancing Taxi Placement in Urban Areas Using Dominating Set Algorithm with Node and Edge Weights</a>

The Billboard magazine is a world renowned music publication since 1984 and it releases a weekly ranking of the top 100 songs in various categories such as rock, pop, hip-hop, etc. Several studies have determined that it is possible to predict the approximate bucket of ranks that a song is likely to chart in using social and subjective indicators. The definitions of these indicators however can change over time, thus rendering the previous classifications erroneous. Here, we report successful results from our experiments in predicting the ranks and the number of weeks the songs are likely to stay on the charts, using objective and well-defined features, obtained from Spotify’s Web API. It extends existing research about classifying songs into rank buckets of Top-10 and Top-40 using these objective features, demonstrating that it is possible to predict exact ranks of the songs within a root-mean-squared-error of 28 ranks and the number of weeks of charting within a root-mean-squared-error of 7 weeks, demonstrating definitive trends between individual features and the ranks of the songs, demonstrating that objective metadata about the songs serve as good indicators about the trends in the Billboard charts, and can be used to predict a song’s performance on the charts within acceptable error rates.

<a href="https://link.springer.com/chapter/10.1007/978-3-031-15512-3_19">Performance Prediction of Songs on Online Music Platforms</a>

The efficacy of Action Recognition methods depends upon relevance of the action (Verb) with respect to the subject (Noun). Existing methods overlook the suitability of Noun-Verb combination in defining an action. In this work, we propose an algorithm called Reduced Verb Set Generator (RVSGen) to reduce the number of possible verbs related to the actions, based upon the relevance of noun-verb combination. A dual modal fusion model for egocentric activity recognition is proposed here to combine the features extracted from the RGB channels and the Optical flow vectors for an egocentric video to recognize the human activity. Unlike state-of-the-art methods, where the key objects and temporal cues are extracted simultaneously, the proposed model first extracts the spatial features i.e., object information (Noun) from the RGB channels and then relates the Noun with the suitable action (Verb) obtained from motion information (Optical Flow). The verbs are predicted by an ConvLSTM architecture with the help of a modified softmax. The notion behind the modified softmax is to estimate the probability distribution with the reduced verb set obtained from RVSGen and the feature vector obtained from the ConvLSTM. With the help of an end-to-end trained architecture, the noun and verb are predicted which are then concatenated together constituting an action. The experiments are performed on a benchmark dataset. The results show the efficacy of the proposed method compared to the state-of-the-art. The codes related to this work can be found at: \url{https://github.com/mpLogics/EgoAR-RVSGen}.

World University root node

IJCNN

First-person Activity Recognition by Modelling Subject-Action Relevance

Recent years have been witnessing significant developments in research in human action recognition based on skeleton data. The graphical representation of the human skeleton, available with the dataset, provides opportunity to apply Graph Convolutional Networks (GCN), to avail efficient analysis of deep spatial-temporal information from the joint and skeleton structure. Most of the current works in skeleton action recognition use the temporal aspect of the video in short-term sequences, ignoring the long-term information present in the evolving skeleton sequence. The proposed long-term Spatio-temporal Contrastive Learning framework for Skeleton Action Recognition uses an encoder-decoder module. The encoder collects deep global-level (long-term) information from the complete action sequence using efficient self-supervision. The proposed encoder combines knowledge from the temporal domain with high-level information of the relative joint and structure movements of the skeleton. The decoder serves two purposes: predicting the human activity and predicting skeleton structure in the future frames. The decoder primarily uses the high-level encodings from the encoder to anticipate the action. For predicting skeleton structure, we extract an even deeper correlation in the Spatio-temporal domain and merge it with the original frame of the video. We apply a contrastive framework in the frame prediction part so that similar actions have similar predicted skeleton structure. The use of the contrastive framework throughout the proposed model helps achieve exemplary performance while employing a self-supervised aspect to the model. We test our model on the NTU-RGB-D-60 dataset and achieve state-of-the-art performance. The codes related to this work are available at: \url{https://github.com/AnshulRustogi/Long-Term-Spatio-Temporal-Framework}.

Long-term Spatio-temporal Contrastive Learning framework for Skeleton Action Recognition

Single image dehazing is a popular research topic among the researchers in computer vision, machine learning, image processing, and graphics. Most of the recent methods for single image dehazing are based upon supervised learning set up. However, supervised methods require annotation of the data, which often makes the dehazing methods biased towards the manual annotation errors. Unsupervised methods are more likely to produce realistic, clear images. However, fewer efforts are found in the literature for single image dehazing in unsupervised set up. We propose an enhanced CycleGAN architecture for Unpaired single image dehazing, with an attention-based transformer architecture embedded in the generator. The proposed transformer comprises three components: 1) A Feature Attention (FA) block combining channel attention and pixel attention mechanism, 2) A Dynamic feature enhancement block for dynamically capturing the spatial structured features and 3) An adaptive mix-up module to preserve the flow of shallow features from downsampling. Experiments on the benchmark datasets show the efficacy of the proposed method. Codes for this work are available in the link: \url{https://github.com/rsjai47/Attention-Based-CycleDehaze}.