A5100653219- Generative Adversarial Networks and Image Synthesis
- Domain Adaptation and Few-Shot Learning
- Advanced Image and Video Retrieval Techniques
- Remote-Sensing Image Classification
- COVID-19 epidemiological studies
- Image Enhancement Techniques
- Visual Attention and Saliency Detection
- Visual perception and processing mechanisms
- Image and Video Quality Assessment
- Face Recognition and Perception
- Natural Language Processing Techniques
- Anomaly Detection Techniques and Applications
- Recommender Systems and Techniques
- Advanced Image Processing Techniques
- Water resources management and optimization
- Biomedical Text Mining and Ontologies
- Perovskite Materials and Applications
- Topic Modeling
- Advanced Vision and Imaging
- Solid-state spectroscopy and crystallography
- Machine Learning in Materials Science
- Advanced Neural Network Applications
- Chalcogenide Semiconductor Thin Films
- Land Use and Ecosystem Services
- Video Surveillance and Tracking Methods
Bennett University
2024-2025
Government Medical College
2024
Indian Institute of Technology Madras
2023
Meril Life Sciences (India)
2023
Institute of Management Technology
2023
Lovely Professional University
2023
Stanford University
2019-2022
Indian Institute of Technology Bombay
2017-2021
Adobe Systems (United States)
2018-2019
Indian Space Research Organisation
2017
For more than 100 years, the fruit fly
Understanding and predicting the human visual attention mechanism is an active area of research in fields neuroscience computer vision. In this paper, we propose DeepFix, a fully convolutional neural network, which models bottom-up via saliency prediction. Unlike classical works, characterize map using various hand-crafted features, our model automatically learns features hierarchical fashion predicts end-to-end manner. DeepFix designed to capture semantics at multiple scales while taking...
Contrastive learning methods have significantly narrowed the gap between supervised and unsupervised on computer vision tasks. In this paper, we explore their application to geo-located datasets, e.g. remote sensing, where unlabeled data is often abundant but labeled scarce. We first show that due different characteristics, a non-trivial persists contrastive standard benchmarks. To close gap, propose novel training exploit spatio-temporal structure of sensing data. leverage spatially aligned...
Convolutional Neural Network(CNN) based semantic segmentation require extensive pixel level manual annotation which is daunting for large microscopic images. The paper aimed towards mitigating this labeling effort by leveraging the recent concept of generative adversarial network(GAN) wherein a generator maps latent noise space to realistic images while discriminator differentiates between samples drawn from database and generator. We extend multi task learning discriminator-classifier...
Accurate local-level poverty measurement is an essential task for governments and humanitarian organizations to track the progress towards improving livelihoods distribute scarce resources. Recent computer vision advances in using satellite imagery predict have shown increasing accuracy, but they do not generate features that are interpretable policymakers, inhibiting adoption by practitioners. Here we demonstrate computational framework accurately at a local level applying object detectors...
Image-based virtual try-on for fashion has gained considerable attention recently. The task requires trying on a clothing item target model image. An efficient framework this is composed of two stages: (1) warping (transforming) the cloth to align with pose and shape model, (2) texture transfer module seamlessly integrate warped onto Existing methods suffer from artifacts distortions in their output. In work, we present Sieve Net, robust image-based try-on. Firstly, introduce multi-stage...
Understanding and predicting the human visual attentional mechanism is an active area of research in fields neuroscience computer vision. In this work, we propose DeepFix, a first-of-its-kind fully convolutional neural network for accurate saliency prediction. Unlike classical works which characterize map using various hand-crafted features, our model automatically learns features hierarchical fashion predicts end-to-end manner. DeepFix designed to capture semantics at multiple scales while...
The combination of high-resolution satellite imagery and machine learning have proven useful in many sustainability-related tasks, including poverty prediction, infrastructure measurement, forest monitoring. However, the accuracy afforded by comes at a cost, as such is extremely expensive to purchase scale. This creates substantial hurdle efficient scaling widespread adoption high-resolution-based approaches. To reduce acquisition costs while maintaining accuracy, we propose reinforcement...
Polymer nanocomposites (PNCs) offer a broad range of thermophysical properties that are linked to their compositions. However, it is challenging establish universal composition-property relationship in PNCs due wide-ranging composition and chemical space. Here, we address this problem develop new method model the composition-microstructure relation PNC through an intelligent machine-learning pipeline named nanoNET. The nanoNET nanoparticles (NPs) distribution predictor, built upon computer...
Although hybrid halide perovskites $(\text{MAPb}{X}_{3},$ $\mathrm{MA}={\mathrm{CH}}_{3}{\mathrm{NH}}_{3} \mathrm{and} X=\mathrm{I}, \mathrm{Br}, \mathrm{Cl})$ have been ubiquitously explored from the photovoltaic perspective, there are still a few unanswered questions which require more fundamental understanding. One such unsettled issue is puzzling behavior of band gap. Unlike conventional semiconductors, $\text{MAPb}{X}_{3}$ $(X=\mathrm{I}, \mathrm{Br})$ found to show blueshift (increase)...
Image-based virtual try-on for fashion has gained considerable attention recently. This task requires to fit an in-shop cloth image on a target model image. An efficient framework this is composed of two stages: (1) warping the align with body shape and pose model, (2) composition module seamlessly integrate warped onto Existing methods suffer from artifacts distortions in their output. In work, we propose use auxiliary learning power existing state-of-the-art network. We leverage prediction...
Almost all the state-of-the-art neural networks for computer vision tasks are trained by (1) pre-training on a large-scale dataset and (2) finetuning target dataset. This strategy helps reduce dependence improves convergence rate generalization task. Although datasets is very useful new methods or models, its foremost disadvantage high training cost. To address this, we propose efficient filtering to select relevant subsets from Additionally, discover that lowering image resolutions in step...
Visual content based product retrieval has become increasingly important for e-commerce. Fashion retrieval, in particular, is a challenging problem owing to wide range of deformations clothing items along with visual distortions their images. In this paper, we propose Grid Search Network (GSN) learning feature embeddings fashion retrieval. The proposed approach posits the training procedure as search problem, focused on locating matches reference query image grid containing both positive and...
The photoluminescence (PL) decay of hybrid halide perovskite single crystals (MAPbX3, MA = CH3NH3+, Pb Pb2+, X Br–, and I–) is measured over 4 orders magnitude in intensity the time scales 100s nanoseconds to a few microseconds. This long PL non-exponential, suggesting presence distribution carrier relaxation times. Spectro-temporal studies show that emission peak red-shifts with increasing time. physics this problem closely related donor–acceptor pair recombination crystalline...
Visual compatibility prediction refers to the task of determining if a set items go well together. Existing techniques for prioritize sensitivity type or context in item representations and evaluate using fill-in-the-blank (FITB) task. We scale FITB stresstest existing methods which highlights need framework that is sensitive multiple modalities relationships. In this work, we introduce unified learning jointly conditioned on type, context, style. The composed TC-GAE, graph-based network...
This study explores the effects of over-the-top content by examining data from popular streaming services such as Netflix, Hotstar Disney Plus, and Amazon Prime in order to learn more about consumer preferences, industry trends, cross-cultural film exchange. To improve user experience, makes use techniques including textual reviews analysis machine learning methods (K-Means Clustering, Linear Regression, Support Vector Machine Regression). Issues with OTT platforms, like churn issue biased...
Despite the proliferation of wearable health trackers and importance sleep exercise to health, deriving actionable personalized insights from data remains a challenge because doing so requires non-trivial open-ended analysis these data. The recent rise large language model (LLM) agents, which can use tools reason about interact with world, presents promising opportunity enable such at scale. Yet, application LLM agents in analyzing personal is still largely untapped. In this paper, we...
Wearable sensors have become ubiquitous thanks to a variety of health tracking features. The resulting continuous and longitudinal measurements from everyday life generate large volumes data; however, making sense these observations for scientific actionable insights is non-trivial. Inspired by the empirical success generative modeling, where neural networks learn powerful representations vast amounts text, image, video, or audio data, we investigate scaling properties sensor foundation...