Automated Federated Deep Learning
Robot-Assisted Diagnosis of Autism Spectrum Disorder
Early detection of Autism Spectrum Disorder (ASD) is crucial for deciding the appropriate educational and behavioral intervention at the most suitable time. However, there are no absolute biological markers for autism and accurate diagnosis of ASD requires extensive training and experience acquired over the years, such expertise is limited to few individuals centered in metropolitans and is beyond the reach of most of the affected population. robot-assisted interventions have found increasing acceptance as a support tool for therapy and education for children with autism (CwA). CwA prefers to interact with technological tools rather than human beings and hence, robot-assisted diagnosis systems can be employed to improve the early detection of ASD in an automated assessment manner making the ASD diagnosis more objective. The overarching goal of this project is to develop a robot-assisted system for the diagnosis of ASD suitable to children of Indian ethnicity. Upon validation, the benefits of this project can be made available to the unreachable children masses of India.
Orbit Computation of Resident Space Objects for Space Situational Awareness
Due to increased human activity in the last two decades, near-earth space is becoming congested from functional/non-functional satellites and space debris. These space objects of human origin along with natural asteroids and space weather possess natural, accidental or intentional threats to the functional and expensive satellites. Currently Space Surveillance Network (SSN) owned by USA provides orbit information for various Resident Space Objects, which are less accurate. Additionally, India depends on Combined Space Operation Centre (CSpOC) for threat alerts from RSO for active space missions. In this project, we are developing a prototype computationally efficient orbit computation software tool for RSO using NSM infrastructure. It is expected that the outcome of the project will contribute towards India's self-reliance in risk assessment for space assets by providing operationally flexible, scalable, transparent, and indigenous collision probability solution.
A Web Server for Drug Response Prediction using scRNA-seq Data
Single cell RNA sequencing (scRNA-seq) technologies have refined our appreciation for intra and inter tumoral heterogeneity across cancer types. While scRNA-seq has been adopted as a method of choice for a large number of clinical studies, we are yet to exploit it fully for making patient-specific treatment decisions by factoring in Intra-Tumoral-Heterogeneity (ITH). To cater to this need, we will develop a prediction webserver to enable systematic interrogation of intratumoral heterogeneity using scRNA-seq data to arrive at subclonal drug response inferences. Researchers will be able to use this webserver to explain drug resistance in PDX models, cell-lines, and patient tumors. Clinicians, who benchmark scRNA-Seq data for treatment decision making, will also use the proposed platform as a Clinical Decision Support System (CDSS) for sub-clonal drug-response analysis.
LEAD-SV: Low-footprint, Efficient and Adaptive Deep Models for Speaker Verification in Smart Home Devices
Learning to listen - approaches that will enable on-device deployment of SV systems. Learning to detect - approaches for novel countermeasures for spoofing attacks
AI for Monitoring of Wildlife for Conservation
Autonomous Last Mile Vehicle (ALIVE)
The goal of this project is to create an autonomous shuttle for the urban Indian last mile. The shuttle will operate on predefined routes and provide hassle free anytime connectivity to modes of public transport like the Metro. To keep the costs per shuttle low, we plan to leverage state-of-the-art Artificial Intelligence and other computational techniques that enable reducing the costs of sensing hardware needed per shuttle. In its first avatar the shuttle will provide connectivity between points in a university campus.
EngageMe: Multimodal Analysis of Attention among Children with Attention Deficit Hyperactivity Disorder for Digital Learning
Children with Specific Learning Disabilities (SLDs) experience repeated failures and poor performance despite their continuous efforts and practice in learning. At the same time, worldwide, the condition with SLDs has been exacerbated due to the COVID-19 pandemic when education delivery shifted online. Thus, strengthening online education delivery will be important and impacting. However, research has indicated that educators might not always be aware of their students' attentional focus, and this may be particularly true for novice teachers. The aim of this project is to develop an AI-empowered tool that will offer personalized, monitored, and evidence-based identification of attention levels among children with SLD. Once validated, the findings of this project can improve and monitor the attention of children with SLDs and can play a significant role in their inclusion during digital learning.