SEARCH · Search NASA

Results for “pixel reliability”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 127 records · Page 7

All sources NASA NTRS DOE OSTI

DOE OSTI2015

Development of an accurate EPID-based output measurement and dosimetric verification tool for electron beam therapy

Purpose: To develop an efficient and robust tool for output measurement and absolute dose verification of electron beam therapy by using a high spatial-resolution and high frame-rate amorphous silicon flat panel electronic portal imaging device (EPID). Methods: The dosimetric characteristics of the EPID, including saturation, linearity, and ghosting effect, were first investigated on a Varian Clinac 21EX accelerator. The response kernels of the individual pixels of the EPID to all available electron energies (6, 9, 12, 16, and 20 MeV) were calculated by using Monte Carlo (MC) simulations, which formed the basis to deconvolve an EPID raw images to the incident electron fluence map. The two-dimensional (2D) dose distribution at reference depths in water was obtained by using the constructed fluence map with a MC simulated pencil beam kernel with consideration of the geometric and structural information of the EPID. Output factor measurements were carried out with the EPID at a nominal source–surface distance of 100 cm for 2 × 2, 3 × 3, 6 × 6, 10 × 10, and 15 × 15 cm{sup 2} fields for all available electron energies, and the results were compared with that measured in a solid water phantom using film and a Farmer-type ion chamber. The dose distributions at a reference depth specific to each energy and the flatness and symmetry of the 10 × 10 cm{sup 2} electron beam were also measured using EPID, and the results were compared with ion chamber array and water scan measurements. Finally, three patient cases with various field sizes and irregular cutout shapes were also investigated. Results: EPID-measured dose changed linearly with the monitor units and showed little ghosting effect for dose rate up to 600 MU/min. The flatness and symmetry measured with the EPID were found to be consistent with ion chamber array and water scan measurements. The EPID-measured output factors for standard square fields of 2 × 2, 3 × 3, 6 × 6, 10 × 10, 15 × 15 cm{sup 2} agreed with film and ion chamber measurements. The average discrepancy between EPID and ion chamber/film measurements was 0.81% ± 0.60% (SD) and 1.34% ± 0.75%, respectively. For the three clinical cases, the difference in output between the EPID- and ion chamber array measured values was found to be 1.13% ± 0.11%, 0.54% ± 0.10%, and 0.74% ± 0.11%, respectively. Furthermore, the γ-index analysis showed an excellent agreement between the EPID- and ion chamber array measured dose distributions: 100% of the pixels passed the criteria of 3%/3 mm. When the γ-index was set to be 2%/2 mm, the pass rate was found to be 99.0% ± 0.07%, 98.2% ± 0.14%, and 100% for the three cases. Conclusions: The EPID dosimetry system developed in this work provides an accurate and reliable tool for routine output measurement and dosimetric verification of electron beam therapy. Coupled with its portability and ease of use, the proposed system promises to replace the current film-based approach for fast and reliable assessment of small and irregular electron field dosimetry.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2014

SU-E-J-235: Varian Portal Dosimetry Accuracy at Detecting Simulated Delivery Errors

Purpose: To use receiver operating characteristic (ROC) analysis to quantify the Varian Portal Dosimetry (VPD) application's ability to detect delivery errors in IMRT fields. Methods: EPID and VPD were calibrated/commissioned using vendor-recommended procedures. Five clinical plans comprising 56 modulated fields were analyzed using VPD. Treatment sites were: pelvis, prostate, brain, orbit, and base of tongue. Delivery was on a Varian Trilogy linear accelerator at 6MV using a Millenium120 multi-leaf collimator. Image pairs (VPD-predicted and measured) were exported in dicom format. Each detection test imported an image pair into Matlab, optionally inserted a simulated error (rectangular region with intensity raised or lowered) into the measured image, performed 3%/3mm gamma analysis, and saved the gamma distribution. For a given error, 56 negative tests (without error) were performed, one per 56 image pairs. Also, 560 positive tests (with error) with randomly selected image pairs and randomly selected in-field error location. Images were classified as errored (or error-free) if percent pixels with γ<κ was < (or ≥) τ. (Conventionally, κ=1 and τ=90%.) A ROC curve was generated from the 616 tests by varying τ. For a range of κ and τ, true/false positive/negative rates were calculated. This procedure was repeated for inserted errors of different sizes. VPD was considered to reliably detect an error if images were correctly classified as errored or error-free at least 95% of the time, for some κ+τ combination. Results: 20mm{sup 2} errors with intensity altered by ≥20% could be reliably detected, as could 10mm{sup 2} errors with intensity was altered by ≥50%. Errors with smaller size or intensity change could not be reliably detected. Conclusion: Varian Portal Dosimetry using 3%/3mm gamma analysis is capable of reliably detecting only those fluence errors that exceed the stated sizes. Images containing smaller errors can pass mathematical analysis, though may be detected by visual inspection. This work was not funded by Varian Oncology Systems. Some authors have other work partly funded by Varian Oncology Systems.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2017

Design of a Dry Dilution Refrigerator for MMC Gamma Detector Arrays

The goal of this LCP is to develop an ultra-high resolution gamma detector based on magnetic microcalorimeters (MMCs) for accurate non-destructive analysis (NDA) of nuclear materials. For highest energy resolution, we will introduce erbium-doped silver (Ag:Er) as a novel sensor material to replace current Au:Er sensors. The detector sensitivity will be increased by developing arrays of 32 Ag:Er pixels read out by 16 SQUID preamplifiers. MMC detectors require operating temperatures of ~15 mK and thus the use of a dilution refrigerator, and the desire for user-friendly operation without cryogenic liquids requires that this refrigerator use pulse-tube pre-cooling to ~4 K. For long-term reliability, we intend to re-design the heat switch that is needed to apply the magnetizing current to the Ag:Er sensor and that used to fail in earlier designs after months of operation. A cryogenic Compton veto will be installed to reduce the spectral background of the MMC, especially at low energies where ultra-high energy resolution is most important. The goals for FY16 were 1) to purchase a liquid-cryogen-free dilution refrigerator and adapt it for MMC operation, and 2) to fabricate Ag:Er-based MMC γ-detectors with improved performance and optimize their response. This report discusses the design of the instruments, and progress in MMC detector fabrication. Details of the MMC fabrication have been discussed in an April 2016 report to DOE.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

DOE OSTI2022

A novel approach for calculating galaxy rotation curves using spaxel cross-correlation and iterative smoothing

ABSTRACT Precise measurements of the internal dynamics of galaxies have proven of great importance for understanding the internal dark matter distribution of galaxies. We present a novel method for measuring the line-of-sight (LOS) velocities across the face of galaxies by cross-correlation of spectral pixels (spaxels) and an iterative method of smoothing. On simulated data the method can accurately recover the input LOS velocities for different types of spectra (absorption-line dominated, emission-line dominated, and differing shapes of the continuum), and can handle stellar population radial gradients. Most important of all, it continues to provide reliable measurements of LOS velocities with reasonable uncertainties even when the spectra are very low signal-to-noise ratio (approaching ∼1), which is a challenge for traditional template-fitting approaches. We apply our method to data from a real MaNGA galaxy as a demonstration and find promising results with good precision. This novel approach can be complementary to existing methods primarily based on template fitting.

79 ASTRONOMY AND ASTROPHYSICS↗

NASA NTRS2018

Improved Detection of Low Stratus and Fog at Dawn from Dual Geostationary (COMS and FY-2D) Satellites

A novel method was proposed to detect low stratus and fog (LSF) at dawn during the summertime using near simultaneous observations from dual geostationary-orbit satellites (GEOs): the Korean Communication, Ocean and Meteorological Satellite (COMS; 128.2 degrees East) and the Chinese Feng-Yun-2D (FY-2D; 86.5 degrees East). The orbital positions of the GEOs provided a large difference (around 46.5 degrees) in the viewing zenith angle (VZA) in the study region (122-132 degrees East, 32.5-42.5 degrees North) and high contrast observations at dawn. Numerical simulations were carried out to derive the optical properties of the LSF and the radiative differences between the GEOs due to both the VZA and spectral response function (SRF). The conventional threshold method, which used the visible reflectance at 0.67 microns (R (sub 0.67)) and the brightness temperature difference between 3.7 microns and 11 microns (BTD (sub 3.7-11), had limitations in detecting LSF at dawn due to weak visible light and fast night-to-day transition. We utilized the observed stereo differences (delta R (sub 0.67), delta BTD (sub 3.7-11) between the two GEOs to improve LSF detection. The dual-satellite observations were verified with ground-based data from 45 stations in South Korea co-located with each GEO pixel. The delta R (sub 0.67) threshold value showed better accuracy (78 percent vs. 67 percent) than the conventional R (sub 0.67) threshold method. In addition, the delta BTD (sub 3.7-11) threshold was better (55 percent vs. 38 percent) than the BTD (sub 3.7-11) threshold. The dual-satellite method allowed more reliable LSF detection using the combination of delta R (sub 0.67) and R (sub 0.67), particularly for LSF without cumuliform or high clouds. Our method is applicable to multiple geostationary satellites for continuous LSF monitoring.

geostationary satellite↗

DOE OSTI2014

Correcting incompatible DN values and geometric errors in nighttime lights time series images

The Defense Meteorological Satellite Program’s Operational Linescan System (DMSP-OLS) nighttime lights imagery has proven to be a powerful remote sensing tool to monitor urbanization and assess socioeconomic activities at large scales. However, the existence of incompatible digital number (DN) values and geometric errors severely limit application of nighttime light image data on multi-year quantitative research. In this study we extend and improve previous studies on inter-calibrating nighttime lights image data to obtain more compatible and reliable nighttime lights time series (NLT) image data for China and the United States (US) through four steps: inter-calibration, geometric correction, steady increase adjustment, and population data correction. We then use gross domestic product (GDP) data to test the processed NLT image data indirectly and find that sum light (summed DN value of pixels in a nighttime light image) maintains apparent increase trends with relatively large GDP growth rates but does not increase or decrease with relatively small GDP growth rates. As nighttime light is a sensitive indicator for economic activity, the temporally consistent trends between sum light and GDP growth rate imply that brightness of nighttime lights on the ground is correctly represented by the processed NLT image data. Finally, through analyzing the corrected NLT image data from 1992 to 2008, we find that China experienced apparent nighttime lights development in 1992-1997 and 2001-2008 respectively and the US suffered from nighttime lights decay in large areas after 2001.

47 OTHER INSTRUMENTATION↗

NASA NTRS2019

High Performance GaN/AlGaN Ultraviolet Avalanche Photodiode Detector Technologies

Detection of ultraviolet (UV) bands provides distinct advantages for NASA, defense, and commercial applications, including increased spatial resolution, small pixel sizes, and large format arrays. AlxGa1-xN semiconductor alloys have attracted great interest for detection in the UV spectral region because of their potential for high optical gain, high sensitivity, and low dark current performance in ultraviolet avalanche photodiodes (UV-APDs). We are developing GaN/AlGaN UV-APDs that demonstrate consistent and reliable UV-APD performance and operation. For these UV detectors we have measured gains of above 5×10(exp 6) and high quantum efficiencies at ~350 nm enabled by a strong avalanche multiplication process. These UV-APDs are fabricated through high quality metal organic chemical vapor deposition (MOCVD) growth on lattice-matched, low dislocation density GaN substrates with optimized GaN/AlGaN UV-APD material growth and doping parameters. The high performance, variable-area GaN/AlGaN UV-APD detectors and arrays can be customized to a wide variety of sizes including large-area formats to enable sensing and high-resolution detection over UV bands of interest.

Sood, Ashok K.↗

DOE OSTI2020

A New Concept for Kilotonne Scale Liquid Argon Time Projection Chambers

We develop a novel Time Projection Chamber (TPC) concept suitable for deployment in kilotonne-scale detectors, with a charge-readout system free from reconstruction ambiguities, and a robust TPC design that reduces high-voltage risks while increasing the coverage of the light-collection system and maximizing the active volume. This novel concept could be used as a far detector module in the Deep Underground Neutrino Experiment (DUNE). For the charge-readout system, we used the charge-collection pixels and associated application-specific integrated circuits currently being developed for the liquid argon (LAr) component of the DUNE Near Detector design, ArgonCube. In addition, we divided the TPC into a number of shorter drift volumes, reducing the total voltage used to drift the ionization electrons, and minimizing the stored energy per TPC. Segmenting the TPC also contains scintillation light, allowing for precise trigger localization and a more expansive light-readout system. Furthermore, the design opens the possibility of replacing or upgrading components. These augmentations could substantially improve the reliability and the sensitivity, particularly for low-energy signals, in comparison to traditional monolithic LArTPCs with projective-wire charge readouts.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

DOE OSTI2015

Dose ratio proton radiography using the proximal side of the Bragg peak

Purpose: In recent years, there has been a movement toward single-detector proton radiography, due to its potential ease of implementation within the clinical environment. One such single-detector technique is the dose ratio method in which the dose maps from two pristine Bragg peaks are recorded beyond the patient. To date, this has only been investigated on the distal side of the lower energy Bragg peak, due to the sharp falloff. The authors investigate the limits and applicability of the dose ratio method on the proximal side of the lower energy Bragg peak, which has the potential to allow a much wider range of water-equivalent thicknesses (WET) to be imaged. Comparisons are made with the use of the distal side of the Bragg peak. Methods: Using the analytical approximation for the Bragg peak, the authors generated theoretical dose ratio curves for a range of energy pairs, and then determined how an uncertainty in the dose ratio would translate to a spread in the WET estimate. By defining this spread as the accuracy one could achieve in the WET estimate, the authors were able to generate lookup graphs of the range on the proximal side of the Bragg peak that one could reliably use. These were dependent on the energy pair, noise level in the dose ratio image and the required accuracy in the WET. Using these lookup graphs, the authors investigated the applicability of the technique for a range of patient treatment sites. The authors validated the theoretical approach with experimental measurements using a complementary metal oxide semiconductor active pixel sensor (CMOS APS), by imaging a small sapphire sphere in a high energy proton beam. Results: Provided the noise level in the dose ratio image was 1% or less, a larger spread of WETs could be imaged using the proximal side of the Bragg peak (max 5.31 cm) compared to the distal side (max 2.42 cm). In simulation, it was found that, for a pediatric brain, it is possible to use the technique to image a region with a square field equivalent size of 7.6 cm{sup 2}, for a required accuracy in the WET of 3 mm and a 1% noise level in the dose ratio image. The technique showed limited applicability for other patient sites. The CMOS APS demonstrated a good accuracy, with a root-mean-square-error of 1.6 mm WET. The noise in the measured images was found to be σ = 1.2% (standard deviation) and theoretical predictions with a 1.96σ noise level showed good agreement with the measured errors. Conclusions: After validating the theoretical approach with measurements, the authors have shown that the use of the proximal side of the Bragg peak when performing dose ratio imaging is feasible, and allows for a wider dynamic range than when using the distal side. The dynamic range available increases as the demand on the accuracy of the WET decreases. The technique can only be applied to clinical sites with small maximum WETs such as for pediatric brains.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2017

SPITZER SECONDARY ECLIPSE DEPTHS WITH MULTIPLE INTRAPIXEL SENSITIVITY CORRECTION METHODS OBSERVATIONS OF WASP-13b, WASP-15b, WASP-16b, WASP-62b, AND HAT-P-22b

We measure the 4.5 μ m thermal emission of five transiting hot Jupiters, WASP-13b, WASP-15b, WASP-16b, WASP-62b, and HAT-P-22b using channel 2 of the Infrared Array Camera (IRAC) on the Spitzer Space Telescope . Significant intrapixel sensitivity variations in Spitzer IRAC data require careful correction in order to achieve precision on the order of several hundred parts per million (ppm) for the measurement of exoplanet secondary eclipses. We determine eclipse depths by first correcting the raw data using three independent data reduction methods. The Pixel Gain Map (PMAP), Nearest Neighbors (NNBR), and Pixel Level Decorrelation (PLD) each correct for the intrapixel sensitivity effect in Spitzer photometric time-series observations. The results from each methodology are compared against each other to establish if they reach a statistically equivalent result in every case and to evaluate their ability to minimize uncertainty in the measurement. We find that all three methods produce reliable results. For every planet examined here NNBR and PLD produce results that are in statistical agreement. However, the PMAP method appears to produce results in slight disagreement in cases where the stellar centroid is not kept consistently on the most well characterized area of the detector. We evaluate the ability of each method to reduce the scatter in the residuals as well as in the correlated noise in the corrected data. The NNBR and PLD methods consistently minimize both white and red noise levels and should be considered reliable and consistent. The planets in this study span equilibrium temperatures from 1100 to 2000 K and have brightness temperatures that require either high albedo or efficient recirculation. However, it is possible that other processes such as clouds or disequilibrium chemistry may also be responsible for producing these brightness temperatures.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2026

Methods development towards automated, physics-informed, quantitative quality control of TRISO-SiC

Tristructural-isotropic (TRISO) fuel particles have been developed as a high-performance fuel for use in high-temperature gas-cooled reactor (HTGR) systems due to their high efficiency and stability under both normal and off-normal conditions. Broader deployment of this technology in advanced nuclear applications may benefit from quantitative quality assurance and quality control (QA/QC) methods that directly link TRISO properties to downstream performance. A key TRISO property is the SiC layer microstructure, which influences fission product retention during irradiation. However, existing QA/QC for the TRISO-SiC microstructure comprises only a qualitative visual inspection; therefore, there is a clear opportunity for the development of quantitative methods for TRISO QA/QC. Here, to this end, previous work has demonstrated an image processing approach to grain boundary (GB) identification and subsequent extraction of microstructural metrics; however, extensive twinning within the SiC layer complicates such analyses because twin GBs significantly influence microstructural metrics but are not expected to contribute to fission product transport. This study presents the initial development, training, and testing of an ML-based image segmentation algorithm designed to identify and remove twin GBs from standard backscattered electron micrographs, providing an industrially applicable, quantitative, and physically meaningful QA/QC approach for the TRISO-SiC microstructure. Although pixel-wise performance metrics for the twin predictions are low, the change in grain area and the number of GB pixels after twin removal predicted by the ML workflow are within 1% of the true values calculated using crystallographic data. This suggests that the model is well capable of predicting overall twin boundary structures and grain morphology, and continued advancement of this approach could enable automated, scalable, and physics-informed QA/QC for TRISO-SiC microstructures, supporting the reliable qualification of coated particle fuels for next-generation reactor systems.

Computer vision↗

DOE OSTI2020

Magnetic Microcalorimeter (MMC) Gamma-Ray Detectors with Ultra-High Energy Resolution. Comprehensive Technology Readiness Assessment

Magnetic Microcalorimeter (MMC) gamma ray detectors have demonstrated very high energy resolution of 38 eV FWHM for gamma ray energies below 100 keV. This is an order or magnitude better than conventional HPGe detector and can remove most line-overlap problems in NDA by gamma spectroscopy below ~130 keV. Unlike superconducting transition edge sensors (TESs), MMCs have a reproducible and almost perfectly linear response, which greatly simplifies adding spectra from different detector pixels and reduces systematic errors in quantitative assays. The instrument is currently at TRL-5 but could advance to TRL-6 by improving the signal processing software and using the system for the quantitative analysis of a safeguards sample provided by e.g. the IAEA. Making MMC detectors commercially viable and used more widely then depends on making the technology more reliable, specifically by increasing the device uniformity and yield, and by developing commercial software. This situation is often only half-jokingly referred to as the “valley of death” and could be addressed though the SBIR program. Our collaborators at STAR Cryoelectronics LLC are well qualified to further advance MMC development, and XIA LLC is well-positioned to develop the data acquisition system and write the software for MMC readout. In addition, the advance to TRL-6 would be desirable to increase the visibility and showcase the value of MMC spectrometers, e.g. by placing one of them into the Safeguards Analytical Laboratory (SAL) at the IAEA, whose NDA experts have recently started to show interest in cryogenic detector technology. As a result of this LCP, we also now have an instrument at LLNL that we will use for various applications of MMCs in nuclear safeguards and science. We are confident that as the number of high-profile results increases, so will the interest in and commercial viability of this technology.

42 ENGINEERING↗

DOE OSTI2020

Magnetic Microcalorimeter (MMC) Gamma-Ray Detectors with Ultra-High Energy Resolution. Comprehensive Technology Readiness Assessment

42 ENGINEERING↗

DOE OSTI2025

Flat TBPS Integration Testing for the CMS Outer Tracker Upgrade

The upcoming High-Luminosity (HL) LHC will significantly increase luminosity, introducing more demanding operating conditions for the CMS detector. To meet these challenges, CMS is undergoing major upgrades, including a complete replacement of its tracking detector. This poster presents recent integration testing efforts conducted at Fermilab on the Flat Tracker Barrel with pixel-strip (PS) modules integration test stand, a subcomponent of the outer tracker located in the central barrel region. In this region, PS modules are mounted on structures that provide mechanical support and cooling, called planks. The testing focuses on validating the integrated system’s thermal and electrical performance. Results from these tests will be discussed, highlighting their importance for ensuring reliable tracker operation in the HL-LHC era.

Salazar Segovia, Itzelli [UC, Davis]↗

DOE OSTI2024

An extreme thermal cycling reliability test of ATLAS ITk Strips barrel modules

At the end of Run 3 of the Large Hadron Collider (LHC), the accelerator complex will be upgraded to the High-Luminosity LHC (HL-LHC) in order to increase the total amount of data provided to its experiments. To cope with the increased rates of data, radiation, and pileup, the ATLAS detector will undergo a substantial upgrade, including a replacement of the Inner Detector with a future Inner Tracker, called the ITk. The ITk will be composed of pixel and strip sub-detectors, where the strips portion will be composed of 17,888 silicon strip detector modules. During the HL-LHC running period, the ITk will be cooled and warmed a number of times from about -35°C to room temperature as part of the operational cycle, including warm-ups during yearly shutdowns. To ensure ITk Strips modules are functional after these expected temperature changes, and to ensure modules are mechanically robust, each module must undergo ten thermal cycles and pass a set of electrical and mechanical criteria before it is placed on a local support structure. This paper describes the thermal cycling Quality Control (QC) procedure, and results from the barrel pre-production phase (about 5% of the production volume). Additionally, in order to assess the headroom of the nominal QC procedure of 10 cycles and to ensure modules don't begin failing soon after, four representative ITk Strips barrel modules were thermally cycled 100 times — this study is also described.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

DOE OSTI2023

A comparison of past and present computational methods for shape analysis of double-shell x-ray radiographs

Implosion symmetry is a key requirement in achieving a robust burning plasma in inertial confinement fusion experiments. In double-shell capsule implosions, we are interested in the shape of the inner shell as it pushes on the fuel. Shape analysis is a popular technique for studying said symmetry during implosion. Combinations of filtering and contour-finding algorithms are studied for their promise in reliably recovering Legendre shape coefficients from synthetic radiographs of double-shell capsules with applied levels of noise. A radial lineout max(slope) method when used on an image pre-filtered with non-local means and a variant of the marching squares algorithm are able to recover p 0 , p 2 , and p 4 maxslope Legendre shape coefficients with mean pixel discrepancy errors of 2.81 and 3.06, respectively, for the noisy synthetic radiographs we consider. Here, this improves upon prior radial lineout methods paired with Gaussian filtering, which we show to be unreliable and whose performance is dependent on input parameters that are difficult to estimate.

47 OTHER INSTRUMENTATION↗

NASA NTRS2022

Unique Observational Constraints on the Seasonal and Longitudinal Variability of the Earth’s Planetary Albedo and Cloud Distribution Inferred From EPIC Measurements

Thorough comparison to observations is key to developing a credible climate model forecasting capability. Deep Space Climate Observatory (DSCOVR) measurements of Earth’s reflected solar and emitted thermal radiation provide a unique observational perspective that permits a more reliable model/data comparison than is possible with the otherwise available satellite data. The uniqueness is in the DSCOVR satellite’s viewing geometry, which enables continuous viewing of the Earth’s sunlit hemisphere from its Lissajous orbit around the Lagrangian L1 point. The key instrument is the Earth Polychromatic Imaging Camera (EPIC), which views the Earth’s sunlit hemisphere with 1024-by-1024-pixel imagery in 10 narrow spectral bands from 317 to 780 nm, acquiring up to 22 high spatial resolution images per day. The additional feature is that the frequency of EPIC image acquisition is nearly identical to that of the climate GCM data generation scheme where climate data for the entire globe are ‘instantaneously’ calculated at 1-h radiation time-step intervals. Implementation of the SHS (Sunlit Hemisphere Sampling) EPIC-view geometry for the in-line GCM output data sampling establishes a precise self-consistency in the space-time data sampling between EPIC observational and GCM output data generation and sampling. The remaining problem is that the GCM generated data are radiative fluxes, while the EPIC measurements are backscatter-dependent radiances. Radiance to flux conversion is a complex problem with no simple way to convert GCM radiative fluxes into spectral radiances. The more expedient approach is to convert the EPIC spectral radiances into broadband radiances by MODIS/CERES-based regression relationships and then into solar radiative fluxes using the CERES angular distribution models. Averaging over the sunlit hemisphere suppresses the meteorological weather noise, but preserves the intra-seasonal larger scale variability. Longitudinal slicing by the Earth’s rotation permits a self-consistent model/data comparison of the longitudinal model/data differences in the variability of the reflected solar radiation. Ancillary EPIC Composite data provide additional cloud property information for climate model diagnostics. Comparison of EPIC-derived seasonal and longitudinal variability of the Earth’s planetary albedo with the GISS ModelE2 results shows systematic overestimate of cloud reflectivity over the Pacific Ocean with corresponding underestimates over continental land areas.

DSCOVR↗

DOE OSTI2016

An analysis of image storage systems for scalable training of deep neural networks

This study presents a principled empirical evaluation of image storage systems for training deep neural networks. We employ the Caffe deep learning framework to train neural network models for three different data sets, MNIST, CIFAR-10, and ImageNet. While training the models, we evaluate five different options to retrieve training image data: (1) PNG-formatted image files on local file system; (2) pushing pixel arrays from image files into a single HDF5 file on local file system; (3) in-memory arrays to hold the pixel arrays in Python and C++; (4) loading the training data into LevelDB, a log-structured merge tree based key-value storage; and (5) loading the training data into LMDB, a B+tree based key-value storage. The experimental results quantitatively highlight the disadvantage of using normal image files on local file systems to train deep neural networks and demonstrate reliable performance with key-value storage based storage systems. When training a model on the ImageNet dataset, the image file option was more than 17 times slower than the key-value storage option. Along with measurements on training time, this study provides in-depth analysis on the cause of performance advantages/disadvantages of each back-end to train deep neural networks. We envision the provided measurements and analysis will shed light on the optimal way to architect systems for training neural networks in a scalable manner.

Lim, Seung-Hwan↗