- Antenna Design and Analysis
- Advanced Antenna and Metasurface Technologies
- Microwave Engineering and Waveguides
- Anomaly Detection Techniques and Applications
- Cell Image Analysis Techniques
- Human Pose and Action Recognition
- Advanced Neural Network Applications
- Identification and Quantification in Food
- Water Quality Monitoring Technologies
- Multimodal Machine Learning Applications
- Advanced Image and Video Retrieval Techniques
- AI in cancer detection
- Image Processing Techniques and Applications
- Image Enhancement Techniques
- Fractal and DNA sequence analysis
- RNA and protein synthesis mechanisms
- Robotics and Sensor-Based Localization
- Digital Holography and Microscopy
- Video Surveillance and Tracking Methods
- Mobile Crowdsensing and Crowdsourcing
- Machine Learning in Bioinformatics
- Digital Imaging for Blood Diseases
- Advanced Optical Imaging Technologies
- Generative Adversarial Networks and Image Synthesis
- Face recognition and analysis
Indian Institute of Science Education and Research Mohali
2024
University of California, Santa Cruz
2014-2021
IBM Research - Almaden
2017-2020
Institute of Technology Management
2015-2020
Cellular Biomedicine Group (United States)
2020
University of California, San Francisco
2017-2019
IBM (United States)
2019
Techno India University
2015
University of Calcutta
2013
Indian Statistical Institute
2010-2011
Controlling higher order modes up to third harmonic of the fundamental operating frequency in a microstrip line-fed patch antenna has been successfully demonstrated. Harmonic rejection achieved at its feed level using highly compact design defected ground structure (DGS). Rejection characteristics have improved adding an open stub line. All possible occurring between and 3rd identified. Relative suppression radiated fields with without DGS-control quantitatively measured effective control...
Pedestrian detection in thermal infrared images poses unique challenges because of the low resolution and noisy nature image. Here we propose a mid-level attribute form multidimensional template, or tensor, using Local Steering Kernel (LSK) as low-level descriptors for detecting pedestrians far images. LSK is specifically designed to deal with intrinsic image noise pixel level uncertainty by capturing local geometry succinctly instead collecting orientation statistics (e.g., histograms HOG)....
The acquisition of increasingly large plankton digital image datasets requires automatic methods recognition and classification. As data size collection speed increases, manual annotation database representation are often bottlenecks for utilization machine learning algorithms taxonomic classification species in field studies. In this paper we present a novel set to perform accurate detection with minimal supervision. Our approach the performance existing supervised when tested on dataset...
One shot, generic object detection involves searching for a single query in larger target image. Relevant approaches have benefited from features that typically model the local similarity patterns. In this paper, we combine (encoded by descriptors) with global context (i.e., graph structure) of pairwise affinities among descriptors, embedding descriptors into low dimensional but discriminatory subspace. Unlike principal components preserve structure feature space, actually seek linear...
In this paper, we propose a graph theoretic technique for recognizing human actions at distance in video by modeling the visual senses associated with poses. The proposed methodology follows bag-of-word approach that starts large vocabulary of poses (visual words) and derives refined compact codebook key using centrality measure connectivity. We introduce “meaningful” threshold on selects each action type. Our contribution includes novel pose descriptor based histogram oriented optical flow...
We propose a generative model for constructing an efficient set of distinctive textures recognizing architectural distortion in digital mammograms. In the first layer proposed two-layer architecture, mammogram is analyzed by multiscale oriented filter bank to form texture descriptor vectorized responses. Our presumes that every can be characterized "bag primitive patterns" and textural primitives (or textons) represented mixture Gaussians which builds up second model. The observed assumed...
In this paper, we propose a graph theoretic approach for recognizing interactions between two human performers present in video clip. We watch primarily the poses of each performer and derive descriptors that capture motion patterns poses. From an initial dictionary (visual words), extract key (or words) by ranking on centrality measure connectivity. argue are nodes which share close semantic relationship (in terms some suitable edge weight function) with all other pose hence said to be...
This work focuses on a specific development aiming towards the detection of human face in an Internet things (IoT) platform. IoT is growing wireless based technology, which connects physical objects supported by different forms electronic hardware. In this work, we propose application meant for security purpose, where faces will be analyzed online. The accepts image person and tries to find match with any pre-stored facial images database. automatically validated status message can...
A recently developed technique to design and model isolated Defected Ground Structure (DGS) has been examined control coupling between two adjacent elements in a microstrip array. This is the only that can handle DGS here, this explored for flrst time antenna domain. An X-band presented. set of prototypes are used obtain measured data which employed verify experimentally
We describe Quantius, a crowd-based image annotation platform that provides an accurate alternative to task-specific computational algorithms for difficult analysis problems. use Quantius quantify variety of computationally challenging medium-throughput tasks with ~50x and 30x savings in time cost respectively, relative single expert annotator. show equivalent deep learning performance Quantius- expert-derived annotations, bridging towards scalable integration tailored machine-learning algorithms.
Plankton is at the bottom of food chain. Microscopic phytoplankton account for about 50% all photosynthesis on Earth, corresponding to 50 billion tons carbon each year, or 125 tonnes sugar[1]. also most species fish, and therefore it represents backbone aquatic environment. Thus, monitoring plankton paramount infer potential dangerous changes ecosystem. In this work we use a collection extracted from large dataset images Woods Hole Oceanographic Institute (WHOI), establish basic set...
One shot, generic object detection involves detecting a single query image in target image. Relevant approaches have benefitted from features that typically model the local similarity patterns. Also important is global matching of along process. In this paper, we consider such information early feature extraction stage by combining geodesic structure (encoded LARK descriptors) with context (i.e., graph structure) pairwise affinities among descriptors. The result an embedding descriptors...
Characterization of defected ground structure (DGS) prior to employing for suppressing mutual coupling between two antennas has been explored the first time. Commercial simulator is used and technique leads physical insight into its operation. Compact open-ring DGS considered as a sample experimentally verified.
Biologists use optical microscopes to study plankton in the lab, but their size, complexity and cost makes widespread deployment of lakes oceans challenging. Monitoring morphology, behavior distribution situ is essential as they are excellent indicators marine environment health provide a majority Earth's oxygen carbon sequestration. Direct in-line holographic microscopy (DIHM) eliminates many these obstacles, image reconstruction computationally intensive produces monochromatic images. By...
The relationship between cellular geometry and state function is apparent, but not yet completely understood. Precise characterization of important in many fields, from pathology to synthetic biology. High-content high-throughput microscopy accessible researchers now more than ever. This allows for collection large amounts images. Naturally, the analysis this data cannot be left manual investigation needs use efficient computing algorithms detection, segmentation, tracking. Annotation...
ABSTRACT The relationship between cellular architecture and state function is apparent, but not yet completely understood. Precise characterization of important in many fields, from pathology to synthetic biology. High-content high-throughput microscopy now more than ever accessible researchers. This allows for collection large amount images. Naturally, the analysis this data cannot be left manual investigation needs resort use efficient computing algorithms detection, segmentation,...
Changes in morphology and swimming dynamics of plankton by exposure to toxic chemicals are studied using a novel new paradigm image acquisition computer vision system. Single cell ciliate Stentor coeruleus enclosed drop water provide means automatically deposit many individual samples on at surface. Chemicals interest added each while the dynamical morphological changes captured with an optical microscope. With techniques, we analyze motion trajectory sample, along its shape information,...
ABSTRACT Plankton is at the bottom of food chain. Microscopic phytoplankton account for about 50% all photosynthesis on Earth, corresponding to 50 billion tons carbon each year, or 125 tonnes sugar[1]. also most species fish, and therefore it represents backbone aquatic environment. Thus, monitoring plankton paramount infer potential dangerous changes ecosystem. In this work we use a collection extracted from large dataset images Woods Hole Oceanographic Institution (WHOI), establish basic...