Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “data classification”

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.

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

At least 1,387 records · Page 77
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Hydra: Computer Vision for Online Data Quality Monitoring

Hydra is a system utilizing computer vision for near real-time data quality monitoring. Currently operational across all of Jefferson Lab’s experimental halls, it reduces the workload of shift takers by autonomously monitoring diagnostic plots during experiments. Hydra uses "off-the-shelf" supervised learning technologies and is supported by a comprehensive MySQL database. To simplify access, web apps have been developed to facilitate both labeling and monitoring of Hydra’s inferences. Hydra can connect with the alarm system and incorporates complete historical tracking, enabling it to identify issues that shift takers could miss. When issues are detected, a natural first question is: "Why does Hydra think there is a problem?" To answer, Hydra employs Gradient-weighted Class Activation Maps (GradCAM) to identify regions of the image that are important for the specific classification. This interpretive layer enhances transparency and trustworthiness, which is essential for integration with experiment workflows and operation. The Hydra system, results, and sociological considerations for deployment will be discussed.

Jeske, Torri↗
DOE OSTI2021

Radiological Anomaly Detection And Identification (RADAI) v1.0

The Radiological Anomaly Detection and Identification (RADAI) software package is a python library for implementing, training, and storing algorithms that detect and identify anomalies in gamma-ray spectra. The library defines a general framework for implementing detection (binary) and identification (classification) algorithms, objects to encapsulate the results of analyses, a variety of temporal filtering tools that can be used in constructing algorithms, and conceptual design that allows easy reading and writing of algorithms (and their time dependent state). In addition to this framework, the library includes implementations of a variety of algorithms from the scientific literature including: gross-counts k-sigma, SPRT, N-SCRAD, Region of Interest, and Censored Energy Window. The implementation of these algorithms within the RADAI package was done to facilitate user-initiated training and configuration to by applied to different gamma-ray detector types. Finally, benchmarked and synthetic datasets will be made available for standardized algorithm characterization with corresponding utilities in the RADAI package for data access and processing.

Joshi, Tenzing↗
DOE OSTI2022

Panel-Segmentation: A Python Package for Automated Solar Array Metadata Extraction Using Satellite Imagery

The NREL Python Panel-Segmentation package is a toolkit that automates the process of extracting accurate and valuable metadata related to solar array installations, using publicly available Google Maps satellite imagery. Previously published work includes automated azimuth estimation for individual solar installations in satellite images. Our continued research focuses on automated detection and classification of solar installation mounting configuration (tracking or fixed-tilt; rooftop, ground, or carport). Specifically, a Faster-RCNN Resnet-50 feature pyramid network (FPN) model was trained and validated on 862 manually labeled satellite images. This model was used to perform object detection on satellite imagery, locating and classifying individual solar installations' mounting configuration and type. Model results showed a mean average precision score (mAP) of 77.79%, with the model strongest at detecting fixed-tilt ground mount and fixed-tilt carport installations. The object detection model and its outputs have been incorporated into the Panel-Segmentation package's automated metadata extraction pipeline, which returns the mounting configuration and azimuth for individual solar arrays in satellite imagery. The complete image data set with labels has been released on the U.S. Department of Energy (DOE) DuraMAT DataHub, to encourage further research in this area.

deep learning↗
DOE OSTI2024

Woody Feedstock 2022 State of Technology Report

The U.S. Department of Energy promotes production of advanced liquid transportation fuels from lignocellulosic biomass by funding fundamental and applied research that advances the state of technology (SOT). As part of its involvement in this mission, Idaho National Laboratory completes an annual SOT report for n th -plant and 1 st -plant woody biomass feedstock logistics. The purpose of the SOT is to provide the status of feedstock supply system technology development for woody biomass to biofuels relative to technical targets and cost goals from specific design cases, based on data and experimental results. Conventional feedstock supply systems need to be modified to meet the demands of conversion pathways, specifically to have the ability to adjust the quality of the raw biomass materials. Advanced systems incorporate innovative methods of material handling, preprocessing and supply chain configuration. In advanced designs, variability of the raw biomass can be reduced to produce feedstocks of a uniform format, moving toward biomass commoditization. Against this backdrop, the 2022 Woody SOT for low-ash woody feedstocks utilizes feedstock fractionation by incorporating technologies that can separate the biomass into its anatomical fractions (wood, bark, needle, and extrinsic ash) to reduce impurities and attempt to maximize the retention of usable fractions that satisfy downstream quality considerations. By using a series of air classification steps, this strategy can reduce the extrinsic ash in forest residues, separate out a majority of the incoming needles (which can be supplied to alternate markets), and maximize the retention of whitewood in the usable fraction. The fractionated forest residues are then mixed with clean-pine chips in a 50-50 blend to prepare the feedstock for the desired conversion pathway. The n th -plant analysis estimated the delivered cost for the feedstock at $\$$69.23/dry ton (2016$\$$) which represents a $\$$6.64/dry ton decrease compared to the cost estimate of the 2021 Woody SOT supply system for low-ash woody feedstocks. The quality requirements in the 2022 Woody SOT were identical to those of the 2021 Woody SOT at = 1.00 wt % ash and = 50.51 wt% carbon. The cost savings derive primarily from reductions in dry matter losses during air classification. The GHG emissions for the n th -plant analysis were estimated at 178.39 kg CO2e/dry ton compared to 178.71 kg CO 2 e/dry ton in the 2021 Woody SOT, a decrease of 0.32 kg CO2e/dry ton. The small change stems from an increase in emissions attributed to preprocessing and slightly larger savings in emissions from transportation. In the 1 st -plant analysis of the 2022 Woody SOT system, the average throughput was estimated to be approximately 2,128 dry tons/day or 96.51% of the name plate capacity. During the simulation the daily throughput ranged from 1,090 dry tons/day to 2,200 dry tons/day, or 49.43% to 99.75% of the daily nameplate capacity. After the year of operation 722,403 tons of processed feedstock were produced in total without regard to quality considerations (99.64% of the annual nameplate capacity). The variability in throughput was primarily caused by equipment failures in the system. Regular failures, downtime caused by routine maintenance per manufacturer guidelines, contributed to a majority 62.50% of failures and 62.60% of downtime. Failures due to wear were the other cause of disruption within the system, impacting the rotary shear and orbital screen and accounting for 37.50% of the failures and 37.40% of the total downtime. Ultimately the system was on stream for 87.84% during the simulation period, which is only 2.16 percentage points below the nth-plant assumption for on-stream time. The production cost of the system averaged $\$$71.66/dry ton. The costs ranged from a minimum of $\$$71.23/dry ton to a maximum of $\$$2,115.30/dry ton. When dry matter losses (disposed low-quality fractions as well as other losses such as in grinders) were considered the costs increased to an average of $\$$75.11/dry ton with a minimum of $\$$74.69/dry ton and a maximum of $\$$2,136.86/dry ton...

09 BIOMASS FUELS↗
DOE OSTI2025

Compositing and Characterization of SE Quadrant Waste Exemplars

This report outlines the experimental investigation and characterization of transport properties in Hanford SE quadrant High-Level Waste (HLW). The goal of the study was to establish baseline behaviors of bulk composite rheology and settling characteristics to facilitate waste treatment process design for the Waste Treatment and Immobilization Plant (WTP) and avoid waste conditions and properties favorable to bubble cascade gas release events. The study focused on two major objectives: 1) identifying, obtaining, and preparing relevant Hanford tank waste samples for evaluation and 2) quantifying the “as-received” rheology and transport properties of the samples. Twenty-three centrifuged core segments originating from tanks AN-101, AN-106, and AW-105 were selected based on compositional relevance to SE quadrant PUREX cladding waste. These materials were composited into five waste composites enriched with target analytes: aluminum (Al), iron (Fe), phosphate (PO 4 ), uranium (U), and zirconium (Zr). Physical property and transport testing examined particle size distributions, bulk densities, settling behaviors, rheological properties, shear strengths, and just-suspended mixing speeds (NJS). Testing revealed two distinct composite classifications based on rheological characteristics: non-Newtonian composites (Fe and PO 4 ) and Newtonian composites (Al, U, and Zr). The Fe and PO 4 composites exhibited slow settling rates and reduced mobilization proclivity, attributable to strong particle-particle interactions and the formation of yield structures within non-Newtonian slurries. In contrast, the Al, U, and Zr composites displayed rapid settling and dense compaction behaviors, indicative of minimal structuring and interactions. Shear strengths for all composites were generally low relative to prior studies of SE quadrant waste, with only the U composite showing elevated strength approaching values reported in previous literature. Repeat shear strength measurements revealed contributions from dense granular material in the U composite and stronger cohesive properties in the Al composite. Settling data highlighted hindered settling behavior, with rates falling more than one order of magnitude below estimates based on Stokes’ law and rate decreasing as composite UDS content increased. NJS testing demonstrated different mobilization behaviors between cohesive and granular composites. The Fe composite required the highest mixing rate for resuspension, while the Al composite was the easiest to resuspend. Comparison of measured NJS against predictions made using the Zwietering correlation suggests non-Newtonian behavior alters resuspension mechanics, rendering non-Newtonian systems more stable against resuspension lift forces relative to their granular counterparts.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2024

Biomass Resources and Emission Reduction Potential of Agricultural and Livestock Residues in Mainland China from 2013 to 2022

Controlling carbon emissions is a global goal, and China is actively implementing carbon reduction measures. As a major agricultural nation, China has considerable potential for developing agricultural residues as renewable and environmentally friendly biomass energy. In this study, we obtained data on crop yields, crop-to-grain ratios, and livestock excretion coefficients to calculate the biomass resources of agricultural and livestock residues in Chinese provinces from 2013 to 2022. Crop residue biomass resources showed a distribution pattern with higher levels in the north than in the south and the east than in the west. Henan and Heilongjiang provinces consistently had the highest resource levels, exceeding 35 million tons annually for 10 years. The biomass resources from livestock residues were relatively abundant in Sichuan, Henan, Yunnan, Shandong, Hunan, and Inner Mongolia. Inner Mongolia, Sichuan, Shandong, and Henan had the greatest potential for CO2 emission reductions, primarily located in regions abundant in biomass resources and with high traditional energy consumption levels. ArcGIS was used to apply natural break classification to categorize the potential for emission reductions from agricultural and livestock residues across China from 2013 to 2022 into five classes. Based on factors such as crop planting area and livestock numbers, the spatiotemporal distribution of factors influencing the quantity of biomass resources was examined using Geographically and Temporally Weighted Regression. A tailored and integrated approach should be used for biomass, and the development of biomass energy should be promoted through policy support and technological innovation.

Environmental Sciences & Ecology↗
DOE OSTI2024

Enhanced Detection of Primary Biological Aerosol Particles Using Machine Learning and Single-Particle Measurement

Accurately identifying primary biological aerosol particles (PBAPs) using analytical techniques poses inherent challenges due to their resemblance to other atmospheric carbonaceous particles. Here, we present a study of an enhanced method for detecting PBAPs by combining single-particle measurement with advanced supervised machine learning (SML) techniques. We analyzed ambient particles from a variety of environments and lab-generated standards, focusing on chemical composition for traditional rule-based and clustering approaches and incorporating morphological features into the SML approaches, neural networks and XGBoost, for improved accuracy. This study demonstrates that SML methods outperform traditional methods in quantifying PBAPs, achieving significant improvements in precision, recall, F1-score, and accuracy, leading to an increased number of detected PBAPs by at least 19%. The adaptability of the proposed XGBoost-based SML model is showcased in comparison to traditional methods in categorizing PBAPs for blind data sets from different geographical locations. Two field case studies were investigated, over agricultural land and Amazonia rain forest, representing relatively low and high concentrations of PBAPs, respectively, where XGBoost consistently detected up to 3.5 times more PBAPs than traditional methods. Precise detection of PBAPs in the atmosphere could significantly improve the prediction of climatic impacts by them.

42 ENGINEERING↗
DOE OSTI2020

Assessment of Spatiotemporal Variability of Evapotranspiration and Its Governing Factors in a Mountainous Watershed at the East River, Colorado, Water: Dataset

This data package contains simulation results of daily evapotranspiration for 17 locations of meteorological stations with different precipitation, air temperature and vegetation within the East River Watershed, Colorado (CLMResult.csv). Daily precipitation data is included for reference. These are the main results of the paper Tran, 2019. Meteorological stations names, latitude and longitude coordinates of meteorological stations, types of lands from the National Land Cover Database, and a comparison with the community land model (CLM) data are also included (Met_LC.csv). The purpose of this data is to analyze the spatio-temporal variations of evapotranspiration at the East River Watershed and understand the factors that control these variations. More information regarding the study, methods, and findings can be found in the Related Reference paper.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Yet Another Discriminant Analysis (YADA): A Probabilistic Model for Machine Learning Applications

This paper presents a probabilistic model for various machine learning (ML) applications. While deep learning (DL) has produced state-of-the-art results in many domains, DL models are complex and over-parameterized, which leads to high uncertainty about what the model has learned, as well as its decision process. Further, DL models are not probabilistic, making reasoning about their output challenging. In contrast, the proposed model, referred to as Yet Another Discriminate Analysis(YADA), is less complex than other methods, is based on a mathematically rigorous foundation, and can be utilized for a wide variety of ML tasks including classification, explainability, and uncertainty quantification. YADA is thus competitive in most cases with many state-of-the-art DL models. Ideally, a probabilistic model would represent the full joint probability distribution of its features, but doing so is often computationally expensive and intractable. Hence, many probabilistic models assume that the features are either normally distributed, mutually independent, or both, which can severely limit their performance. YADA is an intermediate model that (1) captures the marginal distributions of each variable and the pairwise correlations between variables and (2) explicitly maps features to the space of multivariate Gaussian variables. Numerous mathematical properties of the YADA model can be derived, thereby improving the theoretic underpinnings of ML. Validation of the model can be statistically verified on new or held-out data using native properties of YADA. However, there are some engineering and practical challenges that we enumerate to make YADA more useful.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

DMTN-221: Periodicity Analysis in Alert Production

The baselined timeseries features to be computed in Alert Production include a Lomb-Scargle periodogram ran on two classes of variable systems: RR Lyrae and Eclipsing Binaries. Based on a simulated LSST-like cadence light curves taken from the Extended LSST Astronomical Time-series Classification Challenge (ELAsTiCC) we perform an end-to-end test to characterize the periodicity recovery on the Alert Production multi-band light curves. In both variable classes, we found that a single-band Lomb-Scargle implementation yields to a low fraction of recovered periods, with a significant preference on the simple periodic phenomena such as RR Lyrae. We also investigated the results from a multi-band Lomb-Scargle and found an increased fraction of recovered periodicities above 15% for the eclipsing binaries, and over 80$\%$ for the RR Lyrae stars. Our findings suggest that a multi-band Lomb-Scargle should be implemented for searching periodic phenomena through AP. We also asses the computational and scientific performance of several configurations on simulated alert data and find that our current configuration scales linearly with the number of detections while assuming an heuristic frequency grid.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

Impact of Tree Age and Anatomical Fraction on Forest Residue Pyrolysis and Hydrotreating

Utilization of cost-advantaged biomass and waste resources to produce clean, domestic biofuels will be a key factor in decarbonizing transportation in the United States. This study investigates how the distribution of anatomical fractions (stem wood, bark, needles, branches) from pine forest residues of different ages impacts pyrolysis and hydrotreating processes, including product composition, carbon efficiency, economics, and overall sustainability. Commercially sourced 13- and 23-year-old loblolly pine residues were chosen to represent available resources; whole tree thinnings, and tops and branches from merchantable timber, respectively. Detailed characterization of the residue samples, product distributions, bio-oil, char, and gas characterization data are presented for 13 experiments, including the individual fractions, blends, and air-classified samples to reduce ash content. Pyrolysis mass balances averaged 97 +/- 3%. As expected, bio-oil yields were lower for the bark, needles, and blends with high proportions of these (53-56% dry basis), and highest for the debarked stem wood (67-71% dry basis). Lower yields were attributed to the high ash content of the needles, but also to the high lignin content in the bark (>50%), which is hypothesized to have led to increased char formation during pyrolysis. Higher-oil yield was generally correlated with total volatiles, hydrogen, oxygen, glucan, and mannan; somewhat correlated with xylan and acetyl content; and negatively correlated with total carbon, fixed carbon, nitrogen, sulfur, and extractives. Air classification of the residues was effective in reducing ash content (1.5% to 0.64%), extractives (~7% to 2%), and sulfur (0.04% to 0.02%), resulting in a small increase in yield, a significant increase in the GC-detectable proportion of oil (27% to 34%) and a net economic benefit. These experimental results were used to conduct technoeconomic and life cycle analyses, which are also summarized here.

biofuels↗
DOE OSTI2021

DeepLearnMOR: a deep-learning framework for fluorescence image-based classification of organelle morphology

Abstract The proper biogenesis, morphogenesis, and dynamics of subcellular organelles are essential to their metabolic functions. Conventional techniques for identifying, classifying, and quantifying abnormalities in organelle morphology are largely manual and time-consuming, and require specific expertise. Deep learning has the potential to revolutionize image-based screens by greatly improving their scope, speed, and efficiency. Here, we used transfer learning and a convolutional neural network (CNN) to analyze over 47,000 confocal microscopy images from Arabidopsis wild-type and mutant plants with abnormal division of one of three essential energy organelles: chloroplasts, mitochondria, or peroxisomes. We have built a deep-learning framework, DeepLearnMOR (Deep Learning of the Morphology of Organelles), which can rapidly classify image categories and identify abnormalities in organelle morphology with over 97% accuracy. Feature visualization analysis identified important features used by the CNN to predict morphological abnormalities, and visual clues helped to better understand the decision-making process, thereby validating the reliability and interpretability of the neural network. This framework establishes a foundation for future larger-scale research with broader scopes and greater data set diversity and heterogeneity.

Plant Sciences↗
DOE OSTI2023

Energy Optimization of Light and Heavy-Duty Vehicle Cohorts of Mixed Connectivity, Automation and Propulsion System Capabilities via Meshed V2V-V2I and Expanded Data Sharing (Final Scientific and Technical Report)

Vehicle connectivity and automated driving technologies individually have the potential to decrease energy consumption and/or increase safety on light, medium or heavy duty vehicles to varying degrees depending on the traffic infrastructure and specific driving scenarios. Due to advances in sensing, perception and computing power, research and development emphasis in the mobility sector has shifted away from connectivity. Prior research has shown that driving automation with the absence of connectivity can in certain circumstances increase energy consumption. The effectiveness of synergizing connectivity and driving automation technologies is the focus of this work, specifically applied to vehicle cohorts of mixed composition, light and heavy duty, and powertrains ranging from all electric to conventional internal combustion engine. The project team is led by Michigan Technological University (MTU) and partnered with AVL Mobility Technologies Inc. (AVL), Borg Warner (BW), Traffic Technology Services (TTS), American Center for Mobility (ACM) and Navistar (NAV). The main thrusts for the team are to develop a micro-traffic simulation environment with specific VD&PT system attributes and CAV capabilities, 2) field a vehicle test fleet of mixed classification, propulsion and CAV capacity, 3) develop artificial intelligence (AI) and machine learning (ML) based multi-agent optimization methods for various traffic infrastructures, 4) integrate the virtual environment and the optimization methods then deploy the system as a CAV hardware in the loop (HiL) for the vehicle test fleet and 5) conduct closed track and public road testing to validate simulation and demonstrated energy and mobility improvements at multiple scales. For a cohort of mixed vehicles, the team will demonstrate a reduction of energy consumption of 10-50% at intersection, arterial roadway and limited access highway scenarios through connectivity and automation in simulation and at a closed test track. The energy reduction objectives of the project are summarized in Table 1, indicating the infrastructure and over what distances are relevant considered. Single scenario energy reductions are not relevant and thus, the research team took the approach to vary parameters associated with the infrastructure, vehicle cohort composition and dynamic behavior to generate energy consumption distributions for both unconnected and connected scenarios.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2026

Astrometric microlensing probes of the isolated neutron star population with Roman

Notoriously hard to detect and study, isolated neutron stars (NSs) might provide valuable answers to fundamental questions about stellar evolution and explosion physics. With the upcoming Roman Space Telescope, scheduled for launch in 2026, a new and powerful channel for their detection will become available: astrometric microlensing. We set out to create a realistic sample of simulated gravitational microlensing events as observed by Roman with the Galactic Bulge Time Domain Survey. We focus in particular on the population of NS lenses, which has until now been largely understudied. We used dedicated Galactic models tailored for application to microlensing by compact objects. In addition to populations of stars, white dwarfs, and black holes, we simulated four different NS populations with Maxwellian natal kick distributions: v = (150, 250, 350, 450) km/s. For each simulation, we applied projected Roman precision, cadence, and detectability criteria. We found that the parameter space log10 t E –log 10 θ E , which will be accessible to Roman observations, is efficient for the classification of stellar remnants. We found a feature in this space that is characteristic of NSs; using this feature, optimal samples of NS candidates can be constructed from Roman -like datasets. We describe the dependence of the observable parameter distributions on the assumed mean kick velocities. As the effects of natal kicks are very complex and mutually counteracting, we suggest that more detailed studies focused on the dynamics of NSs are needed in anticipation of Roman and future surveys. We estimate that Roman will observe approximately 11 000 microlensing events, including ~100 with NS lenses, whose photometric and astrometric signals are detectable; the event yield decreases by 38% when gap-filling low-cadence observations are not included. We make all simulated microlensing event datasets publicly available in preparation for Roman data.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2025

Understanding EV Charging Pain Points Through Deep Learning Analysis

Current and potential electric vehicle (EV) owners express concerns about the charging infrastructure, mentioning non-functional chargers, prolonged charging times, inconvenient charger locations, long wait times, and high costs as major barriers. Addressing these issues often requires analyzing actual vehicle charging data, which is typically proprietary and inconsistent due to diverse standards and protocols. To understand and improve the EV charging experience, customer reviews are typically used to identify common customer pain points (CPPs). However, there is not a comprehensive method to map customer reviews to a standardized set of CPPs. In collaboration with the National Charging Experience (ChargeX) Consortium, this study bridges these gaps by proposing a Systematic Categorization and Analysis of Large-scale EV-charging Reviews (SCALER) framework. SCALER is an integrated, deep learning framework that segments, actively labels, analyzes, and classifies EV charging customer reviews into six CPP categories. To test its effectiveness, we used SCALER to analyze over 72,000 reviews from customers charging various EV models on different networks across the United States. SCALER achieves a classification accuracy of 92.5%, with an F1 score exceeding 85.7%. By demonstrating real-world applications of SCALER, we enhance the industry’s ability to understand and address CPPs to improve the EV charging experience.

29 - ENERGY PLANNING, POLICY AND ECONOMY↗
DOE OSTI2020

Interactive Visual Study of Multiple Attributes Learning Model of X-Ray Scattering Images

Existing interactive visualization tools for deep learning are mostly applied to the training, debugging, and refinement of neural network models working on natural images. However, visual analytics tools are lacking for the specific application of x-ray image classification with multiple structural attributes. In this paper, we present an interactive system for domain scientists to visually study the multiple attributes learning models applied to x-ray scattering images. It allows domain scientists to interactively explore this important type of scientific images in embedded spaces that are defined on the model prediction output, the actual labels, and the discovered feature space of neural networks. Users are allowed to flexibly select instance images, their clusters, and compare them regarding the specified visual representation of attributes. The exploration is guided by the manifestation of model performance related to mutual relationships among attributes, which often affect the learning accuracy and effectiveness. The system thus supports domain scientists to improve the training dataset and model, find questionable attributes labels, and identify outlier images or spurious data clusters. Case studies and scientists feedback demonstrate its functionalities and usefulness.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Classification of Cloud Particle Imagery and Thermodynamics (COCPIT): A New Databasing Tool for the Characterization of Cloud Particle Images Captured During DOE Field Campaigns

The Department of Energy for decades has explored the earth system and atmosphere through research and deployment of in-situ and remote sensing platforms during field campaigns. Among these datasets exists a vast supply of cloud particle images that provide visual insight into the complex microphysics in the clouds that span our globe. The millions of images collected over decades of deployments provides a unique opportunity to further our understanding of our atmosphere down to the crystal size. This work over the past 5 years has sought to organize these images into digestible datasets that can then be used by scientists to further our understanding of microphysics. A machine learning model was developed that categorizes over 1.5 million images across 11 weather events with over 90% accuracy according to particle type. The database was then extended to include dimensional characteristics of the particle as well as co-location of environmental properties, such as temperature and water content. Then, to initialize the connection between these data and our understanding of how crystals form and grow, weather research and forecasting simulations were run to generate the growth histories of the classified crystals. This research culminates with 2 databases per event: (1) a database of all classified crystals and their dimensional and environmental properties and (2) simulated growth histories of each crystal. Finally, a user interface was created to allow researchers to explore data statistics.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Factorized visual representations in the primate visual system and deep neural networks

Object classification has been proposed as a principal objective of the primate ventral visual stream and has been used as an optimization target for deep neural network models (DNNs) of the visual system. However, visual brain areas represent many different types of information, and optimizing for classification of object identity alone does not constrain how other information may be encoded in visual representations. Information about different scene parameters may be discarded altogether (‘invariance’), represented in non-interfering subspaces of population activity (‘factorization’) or encoded in an entangled fashion. In this work, we provide evidence that factorization is a normative principle of biological visual representations. In the monkey ventral visual hierarchy, we found that factorization of object pose and background information from object identity increased in higher-level regions and strongly contributed to improving object identity decoding performance. We then conducted a large-scale analysis of factorization of individual scene parameters – lighting, background, camera viewpoint, and object pose – in a diverse library of DNN models of the visual system. Models which best matched neural, fMRI, and behavioral data from both monkeys and humans across 12 datasets tended to be those which factorized scene parameters most strongly. Notably, invariance to these parameters was not as consistently associated with matches to neural and behavioral data, suggesting that maintaining non-class information in factorized activity subspaces is often preferred to dropping it altogether. Thus, we propose that factorization of visual scene information is a widely used strategy in brains and DNN models thereof.

59 BASIC BIOLOGICAL SCIENCES↗