IIT Jodhpur | हिंदी संस्करण

 Research

    At present the department has faculties spanning a wide spectrum of Computer Science areas, such as Natural Language processing, Social Network Analysis, AR/VR, Cognitive computing, Image processing, Computer Vision, Artificial Intelligence, Biometrics, IoT, Ad-hoc Wireless Networks, Applied Machine learning, Soft computing and Mobile computing.  
     
 

Dr. Suman Kundu’s research is focused on social networks and its structural aspects. He has employed granular commuting, rough set, and fuzzy sets to solve several network problems such as influence maximization, community detection, link prediction. Currently, he is exploring possibilities of using the social networks or other distributed technologies to solve different tasks for e-Governance, for example, is it possible to trace network activity and predict future events so that a city government can prepare themself for supporting infrastructure and administrative needs on time? Can we process the social media text to know the citizens' needs or how a social scheme is (not)helping the masses?

 
   Suman Kundu  
 

List of completed and on-going projects: 

               
  • The Patterns of Scientific Collaboration with Bibliography Data
  • Event Detection from Social Networks for eGovernance Applications
  • Multi-lingual Text Processing for eGovernance Applications
  • Federated Learning Architecture for Edge devices
  • Blockchain-Based Applications for eGovernance
  • Cyber Bullying in India Social Media
 
 

 

 
 

Dr. Debasis Das’s research interests lie at the intersection of Internet of Things (IoT) and vehicular ad-hoc network (VANET) architectures also termed as Internet of vehicles (IoV). In the future, the IoVs will play an important role in the industry, research organization, and academics for various public and commercial applications. However, the IoV structure should ensure secure and efficient performance for vehicular communications, else an attacker may interfere in the system.

 
 

His proposed novel comprehensive data transmission protocols have the following key features.

               
  • A dependable system with five communications (V2S, V2V, V2I, V2M, and V2R) using one-way hash and elliptic curve operations to achieve mutual authentication between IoV entities (vehicle, RSU, mobile device, wireless sensor, and infrastructure).
  • Resist various security attacks, e.g., Sybil, collision induction, modification, illusion, impersonation, replay, password guessing, man-in-the-middle, and plain-text.
  • Attain better results in different performance measures such as execution time, energy consumption, storage cost, and communication overhead.
 
     
   

List of completed and on-going projects: 

               
  • Secure Vehicular Communication And Routing In Future Intelligent Transportation Systems (ITS)
  • Application of Internet of Vehicles (IoV) For Smart Cities
  • Energy Efficient Communication and Data Flow in Smart City using CRN based IoT Framework
  • IntelliSys: Intelligent System for Road Safety based on Driver Stress and Behavior
  • IoVChain: Blockchain Enabled Secure Communication and Authentication Method for the Internet of vehicles (IoV)
  • Design and Development of Lightweight Cryptographic Solutions for Resource Constrained devices in Smart Cities
 
     
 

Dr. Sumit Kalra’s research interests span multiple areas ranging from Software architecture and Cloud computing to Internet of Things (IoT) and Smart healthcare. He is currently involved in predictive maintenance for industry 4.0 as a method of preventing asset failure by analyzing production data to identify patterns and predict issues proactively. Implementing industrial IoT technologies to monitor asset health, optimize maintenance schedules, and gaining real-time alerts to operational risks, allows manufacturers to lower service costs, maximize uptime, and improve production throughput. The predictive analysis strategies, such as Predicting Remaining Useful Life (RUL), improves performance while maintaining the machine's health. Often the data from heterogeneous sensors is unusable. This data must be preprocessed to extract meaningful information for analytics. The selection of analytical strategy and the corresponding model parameters needed to be identified to get higher accuracy. The overall key objectives are manifold.

 
 
               
  • Collect real-time data from heterogeneous Industrial IoT (IIoT) sensors.
  • Design appropriate data infusion techniques to handle incompatible protocols and data.
  • Real-time processing to ingest terabytes of unreliable, drifted sensor data to generate a reliable predictive analysis.
  • Perform inference and data visualization to help the decision-making process using an interactive dashboard.
 
     
   

List of completed and on-going projects: 

               
  • Predictive Maintenance and Quality Control in industries under Industry 4.0
  • Language Independent Speech Generation System
 
     
 

Dr. Yashaswi Verma works in the area of Computer vision and Applied machine learning. He is specifically interested in multi-label learning, where the objective is to assign a sample to a subset of relevant labels from a fixed vocabulary. In general, this can be viewed as a one-to-many assignment task, and has applications in data annotation, archiving and retrieval.

 
     
   

List of completed and on-going projects: 

               
  • Understanding semantic associations between visual and textual data: What lies ahead?
  • Autism diagnosis using multimodal information
  • extreme Multi-label Learning
 
     
 

Dr. Debarati Chakraborty’s research interests lie in the area of Video processing, Machine learning, Artificial Intelligence and Soft computing. She is presently working on new methodologies for general event precognition from videos in the framework of rough-fuzzy sets. Unsupervised generalized (not deals with some specific events) pre-recognition of events from a video sequence is a challenging issue in video analysis since there is no proper definition available for 'event'.

She is also interested in development of a method of vision-based intelligent mimic by robotic arms. This work enables the controlling of the mechanical motion of robotic arms without employing any mechanical controller. Any change in programming is not required in the case of a pre-programmed automated arm. Above all, the method is entirely unsupervised; that is, no manual intervention/labelled data is required here.

 
   
 
 

 

List of completed and on-going projects: 

               
  • Event precognition from videos with linguistic description
  • Controlling of robotic arm with visual mimic
  • Land-type classification from satellite images
  • Fire intensity prediction from videos
 
     
 

Romi Banerjee works in the area of Natural Language Understanding (NLU) and Cognitive computing. Inspired by Alan Turing's ‘thinking machines’ (1950, Mind), she endeavours to design a cognitive agent that is capable of general contemplation and comprehension, and is largely attuned to the dynamics of real-world social interactions. The agent would ideally derive meaning out of everyday natural language expressions (inclusive of affective components, gestures, idiosyncrasies of speech, etc.) and respond empathetically. Such systems can be envisioned as the foundation of interactive assistive systems (e.g., interactive bots for healthcare scenarios, education-aids for children with reading disorders, semantics-sensitive plagiarism checkers).

 
     
   

List of completed and on-going projects: 

               
  • Dynamic, self-evolving multimodal-multisensory knowledge-graph of procedural (~ commonsense) and declarative information
  • Translation of knowledge-granules into operands of cognitive functions
  • Extraction of- and an algebra of operations on- meanings of colloquial natural language expressions
  • Identification, encoding and representation of novel stimuli (either sensed by the system or descriptions obtained from speaker) at run-time
 
     
 

Ravi Bhandari’s primary research interest lies in the area of Mobile Computing Systems. He is presently interested in investigating the use of mobile technology to several aspects of road safety, viz. “driving monitoring”, “driver monitoring” and “road condition monitoring”. This often involves fusing multiple sensors such as GPS, accelerometer and the camera, and becomes challenging in the presence of noisy real-road conditions. For example, the presence of a small pothole on a road could severely affect the accelerometer reading, even on a “good” quality road. Or location information from GPS may simply be unavailable inside crowded buses or trains. His research is experimental in nature and involves leveraging alternative sensors to setback the limitations of the other, but in a resource constrained environment (such as a smartphone).

 
     
   

List of completed and on-going projects: 

               
  • Road surface quality monitoring
  • Helmetless driving detection
  • Multi-modal room occupancy detection
 
     
 

Chiranjoy Chattopadhyay’s primary research interest lies in the area of Computer Vision. He is presently interested in investigating the use of Computer Vision techniques to several aspects of document images, AR/VR applications in Digital Heritage, Smart Industry Applications like Manufacturing, Food, Game Studies, Digital Humanities. This involves fusing multimodal data, and becomes challenging in the presence of unlabelled data as well as affected by the small sample size problem. His research has applications in the area of real-estate, manufacturing, multimedia, e-commerce, AI.

 
   

 

Fig: Various aspects of applications of Computer Vision techniques for floor plan image analytics.

 

Fig: Various aspects of applications of Computer Vision techniques for smart manufacturing (work done in collaboration with Prof. Kaushal Desai, ME).

 
   

List of completed and on-going projects: 

               
  • Multimodal Floor Plan image retrieval
  • 3D reconstruction for AR/VR based visualization
  • Game Studies 
  • Defect detection and Parameter estimation for Smart Manufacturing Applications
  • Adaptive multimedia streaming for 360 degree live videos
 
     
 

Anand Mishra’s research spans Computer Vision, Language and Knowledge Graphs. To be specific, he focuses on knowledge harvesting from multimodal data and leveraging external knowledge bases for addressing various vision and language tasks.

 
     
 

List of completed and on-going projects: 

               
  • Weakly-supervised visual grounding
  • Sketch-guided object localization
  • Knowledge-enabled Visual Question Answering
  • Interacting with Computer Vision systems in Indian Languages
 
     
 

Deepak Mishra’s research areas include medical image analysis, machine learning, resource constrained AI, biomedical circuits and systems, bio-image computing, and polarization imaging. He is actively involved in developing AI/ML driven healthcare and agricultural solutions while focusing on various aspects of machine learning such as representation learning, lifelong learning, uncertainty modeling, and domain adaptation.

 
     
 

List of completed and on-going projects: 

               
  • Design and development of multimodal fast and safe liver resection device
  • Non-contact, non-invasive SpO2 measurement device using polarization based imaging
  • Content driven on-chip compression for image sensors
  • Effect of latent structure on clustering with GANs