A5091813504- Advanced Neural Network Applications
- CCD and CMOS Imaging Sensors
- Advanced Image and Video Retrieval Techniques
- COVID-19 epidemiological studies
- Industrial Vision Systems and Defect Detection
- Video Surveillance and Tracking Methods
- Domain Adaptation and Few-Shot Learning
- Advanced Memory and Neural Computing
- Indoor and Outdoor Localization Technologies
- Human Pose and Action Recognition
- Anomaly Detection Techniques and Applications
- RFID technology advancements
- Human Mobility and Location-Based Analysis
- Image Retrieval and Classification Techniques
- Video Analysis and Summarization
- Machine Learning and Data Classification
- Autonomous Vehicle Technology and Safety
- Visual Attention and Saliency Detection
- Vehicular Ad Hoc Networks (VANETs)
- Machine Learning and ELM
- Energy Efficient Wireless Sensor Networks
- COVID-19 diagnosis using AI
- Geographic Information Systems Studies
- Educational Management and Quality
- Workplace Spirituality and Leadership
Purdue University West Lafayette
2018-2024
Nvidia (United States)
2023
IEEE Computer Society
2023
Institute of Electrical and Electronics Engineers
2023
Regional Municipality of Niagara
2023
Amrita Vishwa Vidyapeetham
2020
JSS Science and Technology University
2018
PES University
2017
NIIT University
2016
International Institute of Information Technology, Hyderabad
2011-2013
Deep neural networks (DNNs) are successful in many computer vision tasks. However, the most accurate DNNs require millions of parameters and operations, making them energy, computation memory intensive. This impedes deployment large low-power devices with limited compute resources. Recent research improves DNN models by reducing requirement, energy consumption, number operations without significantly decreasing accuracy. paper surveys progress deep learning vision, specifically regards to...
Computer vision has achieved impressive progress in recent years. Meanwhile, mobile phones have become the primary computing platforms for millions of people. In addition to phones, many autonomous systems rely on visual data making decisions, and some these limited energy (such as unmanned aerial vehicles also called drones robots). These batteries, efficiency is critical. This paper serves following two main purposes. First, examine state art low-power solutions detect objects images....
Abstract Context Many engineering organizations are reimplementing and extending deep neural networks from the research community. We describe this process as learning model reengineering. Deep reengineering — reusing, replicating, adapting, enhancing state-of-the-art approaches is challenging for reasons including under-documented reference models, changing requirements, cost of implementation testing. Objective Prior work has characterized challenges development, but yet we know little...
Instance retrieval has emerged as a promising research area with buildings the popular test subject. Given query image or region, objective is to find images in database containing same object scene. There been recent surge efforts finding instances of building challenging datasets such Oxford 5k dataset [19], 100k and Paris [20].
This paper proposes a proof-of-concept for novel automated indoor/outdoor navigation system. Our proposed method shall enable an object/user equipped to be able navigate through closed environments using automatically generated Spatial Map Graph (SMG) with the aid of pre-placed visual markers. The system is robust dynamically changing complex environments, adaptive reconfigurations in SMG during execution. We show that it possible find optimal routes among several interconnected paths...
Embedded devices are generally small, battery-powered computers with limited hardware resources. It is difficult to run deep neural networks (DNNs) on these devices, because DNNs perform millions of operations and consume significant amounts energy. Prior research has shown that a considerable number DNN’s memory accesses computation redundant when performing tasks like image classification. To reduce this redundancy thereby the energy consumption DNNs, we introduce Modular Neural Network...
Computer vision on low-power edge devices enables applications including search-and-rescue and security. State-of-the-art computer algorithms, such as Deep Neural Networks (DNNs), are too large for inference devices. To improve efficiency, some existing approaches parallelize DNN across multiple How-ever, these techniques introduce significant communication synchronization overheads or unable to balance workloads This paper demonstrates that the hierarchical architecture is well suited...
We consider the problem of energy-efficient distributed detection to infer presence a target in wireless sensor network and analyze its robustness modeling uncertainties. The sensors make noisy observations target's signal power, which follows isotropic power-attenuation model. Binary local decisions are transmitted fusion center, where global inference regarding is made, based on counting rule. uncertain knowledge of: 1) decay exponent medium; 2) power attenuation constant; 3) distance...
In the blooming era of smart edge devices, surveillance cameras have been deployed in many locations. Surveillance are most useful when they spaced out to maximize coverage an area. However, deciding where place is NP-hard problem and researchers proposed heuristic solutions. Existing work does not consider a significant restriction computer vision: order track moving object, object must occupy enough pixels. The number pixels depends on factors (How far away object? What camera resolution?...
In order to contain the COVID-19 pandemic, countries around world have introduced social distancing guidelines as public health interventions reduce spread of disease. However, monitoring efficacy these at a large scale (nationwide or worldwide) is difficult. To make matters worse, traditional observational methods such in-person reporting dangerous because observers may risk infection. A better solution observe activities through network cameras; this approach scalable and can stay in safe...
Industries keep a check on all statistics of their business and process this data using various mining techniques to measure profit trends, revenue, growing markets interesting opportunities invest. These statistical records increasing increase very fast. Unfortunately, as the grows it becomes tedious task such large set extract meaningful information. Also if generated is in formats, its processing possesses new challenges. Owing size, big stored Hadoop Distributed File System (HDFS). In...
Recent research has focused on Deep Neural Networks (DNNs) implemented directly in hardware. However, larger DNNs require significant energy and area, thereby limiting their wide adoption. We propose a novel DNN quantization technique corresponding hardware solution, CompactNet that optimizes the use of resources even further, through dynamic allocation memory for each parameter. Experimental results MNIST CIFAR-10 datasets, show reduces requirement by over 80%, 12-fold, area 7-fold, when...
Computer vision has achieved impressive progress in recent years. Meanwhile, mobile phones have become the primary computing platforms for millions of people. In addition to phones, many autonomous systems rely on visual data making decisions and some these limited energy (such as unmanned aerial vehicles also called drones robots). These batteries efficiency is critical. This article serves two main purposes: (1) Examine state-of-the-art low-power solutions detect objects images. Since...
Computer vision is often performed using Convolutional Neural Networks (CNNs). CNNs are compute-intensive and challenging to deploy on power-contrained systems such as mobile Internet-of-Things (IoT) devices. computeintensive because they indiscriminately compute many features all pixels of the input image. We observe that, given a computer task, images contain that irrelevant task. For example, if task looking for cars, in sky not very useful. Therefore, we propose CNN be modified only...
Many engineering organizations are reimplementing and extending deep neural networks from the research community. We describe this process as learning model reengineering. Deep reengineering - reusing, reproducing, adapting, enhancing state-of-the-art approaches is challenging for reasons including under-documented reference models, changing requirements, cost of implementation testing. In addition, individual engineers may lack expertise in software engineering, yet teams must apply...
Deep Neural Networks (DNNs) achieve state-of-the-art accuracy in many computer vision tasks, such as object counting. Object counting takes two inputs: an image and query reports the number of occurrences queried object. To high accuracy, DNNs require billions operations, making them difficult to deploy on resource-constrained, low-power devices. Prior work shows that a significant DNN operations are redundant can be eliminated without affecting accuracy. reduce these redundancies, we...
Wireless Sensor Network has been one of the most diversified and widely used area a vast range applications in almost every field. is domain with large number motes that collects data from surrounding after processing, it transfers to sink node through intermediate sensor which finally transmitted base station. With wide applications, comes up major issues i.e. Energy Consumption. Due dense network topology WSN, communication short incurs redundancy sensed data. To reduce this redundancy,...
COVID-19 has resulted in a worldwide pandemic, leading to "lockdown" policies and social distancing. The pandemic profoundly changed the world. Traditional methods for observing these historical events are difficult because sending reporters areas with many infected people can put reporters' lives danger. New technologies needed safely responses policies. This paper reports using thousands of network cameras deployed purpose witnessing activities response continuously provide real-time...
Processing visual data on mobile devices has many applications, e.g., emergency response and tracking. State-of-the-art computer vision techniques rely large Deep Neural Networks (DNNs) that are usually too power-hungry to be deployed resource-constrained edge devices. Many improve DNN efficiency of DNNs by compromising accuracy. However, the accuracy these cannot adapted for diverse applications with different hardware constraints requirements. This paper demonstrates a recent, efficient...
Low-power computer vision on embedded devices has many applications. This paper describes a low-power technique for the object re-identification (reID) problem: matching query image against gallery of previously-seen images. State-of-the-art techniques rely large, computationally-intensive Deep Neural Networks (DNNs). We propose novel hierarchical DNN architecture that uses attribute labels in training dataset to perform efficient reID. At each node hierarchy, small identifies different...
As the complexity of System on Chip(SOC) designs is increasing day by day, verification becoming a complex task to attain. A SOC design consists various intellectual property cores (IP). To verify so many IPs, testbench has be developed which not an easy achieve. So make task, Verification Intellectual Property (VIP) are developed. In this paper, Controller Area Network (CAN) VIP proposed. This using SystemVerilog based universal methodology (UVM. The test environment verified running...
This article analyzes visual data captured from five countries and three U.S. states to evaluate the effectiveness of lockdown policies for reducing spread COVID-19.The main challenge is scale: nearly six million images are analyzed observe how people respond policy changes.
Object detectors are vital to many modern computer vision applications. However, even state-of-the-art object not perfect. On two images that look similar human eyes, the same detector can make different predictions because of small image distortions like camera sensor noise and lighting changes. This problem is called inconsistency. Existing accuracy metrics do properly account for inconsistency, work in this area only targets improvements on artificial distortions. Therefore, we propose a...