A5076249543- Organic Electronics and Photovoltaics
- Conducting polymers and applications
- Human Pose and Action Recognition
- Hand Gesture Recognition Systems
- Advanced Image and Video Retrieval Techniques
- Video Surveillance and Tracking Methods
- Advanced Neural Network Applications
- Perovskite Materials and Applications
- IoT-based Smart Home Systems
- 3D Shape Modeling and Analysis
- Context-Aware Activity Recognition Systems
- Domain Adaptation and Few-Shot Learning
- Multimodal Machine Learning Applications
- Thin-Film Transistor Technologies
- Face and Expression Recognition
- Water Quality Monitoring Technologies
- IoT and Edge/Fog Computing
- Gait Recognition and Analysis
- Semiconductor Lasers and Optical Devices
- Face recognition and analysis
- Advanced Vision and Imaging
- Natural Language Processing Techniques
- Biometric Identification and Security
- Advanced Photocatalysis Techniques
- Diabetic Foot Ulcer Assessment and Management
University of California, Santa Barbara
2025
LNM Institute of Information Technology
2015-2024
Pennsylvania State University
2024
University of Asmara
2024
Chaudhary Sarwan Kumar Himachal Pradesh Krishi Vishvavidyalaya
2023-2024
National Institute of Technology Hamirpur
2024
Indian Institute of Technology Kanpur
2024
Tata Medical Center
2024
National Yang Ming Chiao Tung University
2023
National Environmental Engineering Research Institute
2021-2023
This paper presents a general multi-view feature extraction approach that we call Generalized Multiview Analysis or GMA. GMA has all the desirable properties required for cross-view classification and retrieval: it is supervised, allows generalization to unseen classes, kernelizable, affords an efficient eigenvalue based solution applicable any domain. exploits fact most popular supervised unsupervised techniques are of special form quadratic constrained program (QCQP), which can be solved...
This paper presents a novel way to perform multi-modal face recognition. We use Partial Least Squares (PLS) linearly map images in different modalities common linear subspace which they are highly correlated. PLS has been previously used effectively for feature selection show both theoretically and experimentally that can be across modalities. also formulate generic intermediate comparison framework Surprisingly, we achieve high performance using only pixel intensities as features....
Monocular 3D human-pose estimation from static images is a challenging problem, due to the curse of dimensionality and ill-posed nature lifting 2D-to-3D. In this paper, we propose Deep Conditional Variational Autoencoder based model that synthesizes diverse anatomically plausible 3D-pose samples conditioned on estimated 2D-pose. We show CVAE-based sample set consistent with 2D-pose helps tackling inherent ambiguity in 2D-to-3D lifting. two strategies for obtaining final pose- (a)...
We present a novel learning-based approach for computing correspondences between non-rigid 3D shapes. Unlike previous methods that either require extensive training data or operate on handcrafted input descriptors and thus generalize poorly across diverse datasets, our is both accurate robust to changes in shape structure. Key method feature-extraction network learns directly from raw geometry, combined with regularized map extraction layer loss, based the functional representation....
A 32 nm logic technology for high performance microprocessors is described. 2 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">nd</sup> generation high-k + metal gate transistors provide record drive currents at the tightest pitch reported any or 28 technology. NMOS are 1.62 mA/um Idsat and 0.231 Idlin 1.0 V 100 nA/um I <sub xmlns:xlink="http://www.w3.org/1999/xlink">off</sub> . PMOS 1.37 0.240 The impact of SRAM cell array size on Vccmin reported.
We present a novel method for computing correspondences across 3D shapes using unsupervised learning. Our computes non-linear transformation of given descriptor functions, while optimizing global structural properties the resulting maps, such as their bijectivity or approximate isometry. To this end, we use functional maps framework, and build upon recent FMNet architecture Unlike that approach, however, show learning can be done in purely \emph{unsupervised setting}, without having access...
We present a simple and efficient method for refining maps or correspondences by iterative upsampling in the spectral domain that can be implemented few lines of code. Our main observation is high quality obtained even if input are noisy encoded small number coefficients basis. show how this approach used conjunction with existing initialization techniques across range application scenarios, including symmetry detection, map refinement complete shapes, non-rigid partial shape matching...
Convolutional Neural Networks (ConvNets) have shown excellent results on many visual classification tasks. With the exception of ImageNet, these datasets are carefully crafted such that objects well-aligned at similar scales. Naturally, feature learning problem gets more challenging as amount variation in data increases, models to learn be invariant certain changes appearance. Recent ImageNet dataset show given enough data, ConvNets can invariances producing very discriminative features [1]....
We propose a modular approach towards large-scale real-time object detection by decoupling objectness and classification. exploit the fact that many classes are visually similar share parts. Thus, universal detector can be learned for class-agnostic followed fine-grained classification using (non)linear classifier. Our is modification of R-FCN architecture to learn shared filters performing localization across different classes. trained 3000 classes, called R-FCN-3000, obtains an mAP 34.9%...
This paper proposes a learning-based approach to scene parsing inspired by the deep Recursive Context Propagation Network (RCPN). RCPN is feed-forward neural network that utilizes contextual information from entire image, through bottom-up followed top-down context propagation via random binary parse trees. improves feature representation of every super-pixel in image for better classification into semantic categories. We analyze and propose two novel contributions further improve model....
In recent years, yoga has become part of life for many people across the world. Due to this there is need scientific analysis y postures. It been observed that pose detection techniques can be used identify postures and also assist perform more accurately. Recognition posture a challenging task due lack availability dataset detect on real-time bases. To overcome problem large created which contain at least 5500 images ten different tf-pose estimation Algorithm draws skeleton human body...
An overview of the significant innovations in photocatalysts for H2 development, photocatalyst selection criteria, and photocatalytic modifications to improve activity was examined this Review, as well mechanisms thermodynamics. A variety semiconductors have been a structured fashion, such TiO2-, g-C3N4-, graphene-, sulfide-, oxide-, nitride-, oxysulfide-, oxynitrides, cocatalyst-based photocatalysts. The techniques enhancing compatibility metals nonmetals is discussed order boost...
Naphthalene diimide–bithiophene P(NDI2OD-T2) is a well-known donor–acceptor polymer, previously explored as n-type material in all-polymer solar cells (all-PSCs) and organic field effect transistor (OFETs) applications. The optical, bulk, electrochemical, semiconducting properties of polymer were tuned via random incorporation perylene diimide (PDI) coacceptor with naphthalene (NDI). Three copolymers containing 2,2′-bithiophene donor unit varying compositions (NDI) (xPDI, x = 15, 30, 50 mol...
The best performing methods for 3D human pose estimation from monocular images require large amounts of in-the-wild 2D and controlled annotated datasets which are costly sophisticated systems to acquire. To reduce this annotation dependency, we propose Multiview-Consistent Semi Supervised Learning (MCSS) framework that utilizes similarity in information unannotated, uncalibrated but synchronized multi-view videos motions as additional weak supervision signal guide regression. Our applies...
In recent years, intelligent sensing has gained significant attention because of its autonomous decision-making ability to solve complex problems. Today, smart sensors complement and enhance the capabilities human beings have been widely embraced in numerous application areas. Artificial intelligence (AI) made astounding growth domains natural language processing, machine learning (ML), computer vision. The methods based on AI enable a learn monitor activities by source information real-time...
Abstract It is highly essential that municipal wastewater treated before its discharge and reuse in order to meet the standard requirements for safe marine life farming industries. beneficial use reclaimed water, since availability of fresh water inadequate. An investigation was conducted on Jamnagar Municipal Corporation Sewage Treatment Plant (JMC-STP) develop a feedforward artificial neural network (FF-ANN) model. an alternate modelling/ prediction JMC-STP circumvent over versatile...