- Local Clustering with Mean Teacher for Semi-supervised learning , 2020 25TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION (ICPR) (2021)
- Perceptual metric learning for video anomaly detection , MACHINE VISION AND APPLICATIONS (2021)
- Anomalous cluster detection in spatiotemporal meteorological fields , STATISTICAL ANALYSIS AND DATA MINING (2019)
- Deformable Part Models for Complex Object Detection in Remote Sensing Imagery , BIGSPATIAL 2018: PROCEEDINGS OF THE 7TH ACM SIGSPATIAL INTERNATIONAL WORKSHOP ON ANALYTICS FOR BIG GEOSPATIAL DATA (BIGSPATIAL-2018) (2018)
- FUTURES-DPE: Towards Dynamic Provisioning and Execution of Geosimulations in HPC environments , 26TH ACM SIGSPATIAL INTERNATIONAL CONFERENCE ON ADVANCES IN GEOGRAPHIC INFORMATION SYSTEMS (ACM SIGSPATIAL GIS 2018) (2018)
- Machine Learning Approaches for Slum Detection Using Very High Resolution Satellite Images , 2018 18TH IEEE INTERNATIONAL CONFERENCE ON DATA MINING WORKSHOPS (ICDMW) (2018)
- Real-Time Energy Audit of Built Environments: Simultaneous Localization and Thermal Mapping , JOURNAL OF INFRASTRUCTURE SYSTEMS (2018)
- Hierarchical change detection framework for biomass monitoring , 2017 ieee international geoscience and remote sensing symposium (igarss) (2017)
- High performance GPU computing based approaches for oil spill detection from multi-temporal remote sensing data , Remote Sensing of Environment (2017)
- Parallel processing over spatial-temporal datasets from geo, bio, climate and social science communities: A research roadmap , 2017 ieee 6th international congress on big data (bigdata congress 2017) (2017)
The Science and Technologies for Phosphorus Sustainability (STEPS) Center is a convergence research hub for addressing the fundamental challenges associated with phosphorus sustainability. The vision of STEPS is to develop new scientific and technological solutions to regulating, recovering and reusing phosphorus that can readily be adopted by society through fundamental research conducted by a broad, highly interdisciplinary team. Key outcomes include new atomic-level knowledge of phosphorus interactions with engineered and natural materials, new understanding of phosphorus mobility at industrial, farm, and landscape scales, and prioritization of best management practices and strategies drawn from diverse stakeholder perspectives. Ultimately, STEPS will provide new scientific understanding, enabling new technologies, and transformative improvements in phosphorus sustainability.
In many real-world applications, data loses its value if itâ€™s not analyzed in near real time. Examples include natural disasters, crop disease identification and bioterrorism, traffic monitoring, monitoring human activities and public places. Edge computing refers to pushing computing power to the edge of the network or bringing it closer to the sensors. We envision that the embedded supercomputers (e.g., Jetson TX1 and TX2; 1 Teraflop; ~10 Watts) allow computing at the edge (e.g., UAVs). This framework would then allow near real-time analytics on streaming data, which is critical for first responders to national security agencies alike, and compress/reduce data before transmitted to the cloud or data centers. In this project, we propose to develop novel machine learning algorithms on the embedded supercomputers while the data is still in device memory and demonstrate the technology in two real-world applications: crop monitoring and traffic monitoring. Proposed technical work involves following three key stages. (i) Generate a statistical model from historical data (e.g., spectral signatures of different crops) by using statistically principled mixture model (e.g., Gaussian Mixture Model (GMM)), (ii) As the data is being acquired compare new (streaming) data with the GMM model to identify any anomalous patterns (e.g., weeds), (iii) generate event signal about the anomaly before the data is being compressed and transferred out from devise memory.
NC State University, in partnership with University of Michigan, Purdue University, University of Illinois at Urbana Champaign, Kansas State University, Georgia Institute of Technology, NC A&T State University, Los Alamos National Lab, Oak Ridge National Lab, and Pacific Northwest National lab, proposes to establish a Consortium for Nonproliferation Enabling Capabilities (CNEC). The vision of CNEC is to be a pre-eminent research and education hub dedicated to the development of enabling technologies and technical talent for meeting the grand challenges of nuclear nonproliferation in the next decade. CNEC research activities are divided into four thrust areas: 1) Signatures and Observables (S&O); 2) Simulation, Analysis, and Modeling (SAM); 3) Multi-source Data Fusion and Analytic Techniques (DFAT); and 4) Replacements for Potentially Dangerous Industrial and Medical Radiological Sources (RDRS). The goals are: 1) Identify and directly exploit signatures and observables (S&O) associated with special nuclear material (SNM) production, storage, and movement; 2) Develop simulation, analysis, and modeling (SAM) methods to identify and characterize SNM and facilities processing SNM; 3) Apply multi-source data fusion and analytic techniques to detect nuclear proliferation activities; and 4) Develop viable replacements for potentially dangerous existing industrial and medical radiological sources. In addition to research and development activities, CNEC will implement educational activities with the goal to develop a pool of future nuclear non-proliferation and other nuclear security professionals and researchers.
The PIâ€™s unique combination of expertise in large scale spatial and spatiotemporal data management, data mining, machine learning, and high-performance analytics is necessary to ensure the success of ORNL-based projects in the area of biomass monitoring, rapid damage assessments, human settlement mapping, very high-resolution, multi-sensor remote sensing data analysis, climate change, urban informatics, material genomics and imaging, and national security sponsored by various federal agencies, donors, international institutes, and LDRD programs. The PI will direct and conduct fundamental research, collaborate with research staff, direct post-doc and post-master students, write up research results for peer-review publications, give demonstrations and presentations, and mentor graduate students sponsored through ORNL grants.
The purpose of the updated and extended project is to provide additional technical support, consultation and research related to developing and evaluating new methods in the application of geospatial analysis for the Rivers, Trails, and Conservation Assistance (RTCA) program within the Conservation and Outdoor Recreation (COR) Branch of the National Park Service (NPS). The additional technical support, consultation and research activities will include, but are not limited to: 1) developing training material for using the RTCA web mapping application; 2) incorporating additional COR Branch program data into the existing RTCA Enterprise database; and 3) enhancing the current RTCA web mapping application by incorporating existing themed GIS web services.
Massive amounts of remote sensing data are being collected and archived from satellites and airborne platforms (including drones) on daily basis. This data supports a wide range of applications of national importance. Examples of applications include crop type mapping, forest mapping, urban neighborhood mapping, damages due to flooding, hailstorms, and forest fires, impacts of climate change on crops, unusual crop detection (e.g., poppy plantations), changes in biomass, understanding complex interaction between food, energy, and water, etc. Classification of these high-resolution images requires object and arbitrary patch based classification to capture relevant spatial context. The advent of multiple instance learning and deep learning took the natural image processing community by storm. However, its application to satellite images has been slow due to training data and computational requirements. In this project, we develop deep learning algorithms for classification of satellite images and scale these algorithms on Lenovo/Intelâ€™s new architectures and software infrastructure (e.g., Neon, Caffe, Theano, and MXNet).
Slums have become an inescapable feature of cities in the developing world, and the number of people living in slums has increased rapidly, coming close to 1 billion and rising higher (UN-Habitat 2010). Relatively little is known, however, about patterns of slum development over periods of time and about factors associated with progressive improvements. One of the objectives of this research is to develop a prototype methodology for semi-automatic slum identification and categorization that can speedily and reliably be adapted for use in other cities.
Datasets being generated by experiment and simulation today are increasingly large, and nations across the world â€“ including China, the United States, Europe, and Japan â€“ have all invested heavily in developing computers capable of processing or generating these datasets. These datasets come from applications in many areas, and are driven by national security issues as well as industries of strategic value. Large-scale computing is seen as driving technological developments in biology and biomedicine, high-energy physics (a key to stockpile stewardship), and materials science. All of these are areas where the United States has traditionally led the world. However, recent developments have placed China as the leader in building large-scale computers and there is a concern that this could result in the loss of a leadership role for the United States in many of the related technologies. With this in mind, the United States has placed a renewed emphasis on developing exascale computer platforms, and strategically, on the development of algorithms which can make use of large computers to support decisions in science and engineering â€“ an area where the United States arguably still leads the world. In this proposal, we propose adapting Kitwareâ€™s Catalyst and Cinema platforms to perform new summarization tasks including compression scalably on these new architectures. To demonstrate the effectiveness of these summarizations for this phase I project, we will adapt a simulation program to use Catalyst for in-situ processing â€“ saving only the dynamic summarization â€“ in order to provide stakeholders with the information necessary to make a decision and be confident in the simulation process.
Scaling-up scientific data analysis and machine learning algorithms for data-driven discovery is a challenging task. Despite the growing need for analysis from science domains that are generating â€˜Big Dataâ€™ from instruments and simulations, building high-performance analytical workflows of data-specific algorithms has been an impediment due to: (i) evolving nature of the â€˜Big Dataâ€™ hardware and software architecture landscape, (ii) newer architectures impose new programming models, and (iii) lack of understanding of data-parallel kernels of analysis algorithms and their performance on different architectures and programming environments. NCUS will conduct research on benchmarking core graph kernels and computing primitives.
The purpose of this project is to facilitate the development of a unified framework for the NPS Conservation and Outdoor Recreation Division and related programs to guide establishment of a geospatial platform and consistent approach to the collection, organization, display, and communication of program and project data/information from across the breadth of work supported by NPS outside park units. This collaborative effort will engage and have direct utility for decision makers in the NPS, in partner non-profit organizations, and in state and local government.