SEARCH · Engineering Papers

Results for “IMAGE”

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 37 records · Page 2

All sources NASA NTRS DOE OSTI

DOE OSTI2025

Heliostat optical error inspection with polarimetric imaging drone

On a Concentrated Solar Power (CSP) field, optical errors have significant impacts on the collection efficiency of heliostats. Fast, cost-effective, labor-efficient, and non-intrusive autonomous field inspection remains a challenge. Approaches using imaging drone, i.e., Unmanned Aerial Vehicle (UAV) system integrated with high resolution visible imaging sensors, have been developed to address these challenges; however, these approaches are often limited by insufficient imaging contrast. Here, in this study, we report a polarimetry-based method with a polarization imaging system integrated on UAV to enhance imaging contrast for in-situ detection of heliostat mirrors without interrupting field operation. We developed an optical model for skylight polarization pattern to simulate the polarization images of heliostat mirrors and obtained optimized waypoints for polarimetric imaging drone flight path to capture images with enhanced contrast. The polarimetric imaging-based method improved the success rate of edge detections in scenarios which were challenging for mirror edge detection with conventional imaging sensors. We have performed field tests to achieve significantly enhanced heliostat edge detection success rate and investigate the feasibility of integrating polarimetric imaging method with existing imaging-based heliostat inspection methods, i.e., Polarimetric Imaging Heliostat Inspection Method (PIHIM). Our preliminary field test results suggest that the PIHIM hold the promise to enable sufficient imaging contrast for real-time autonomous imaging and detection of heliostat field, thus suitable for non-interruptive fast CSP field inspection during its operation.

CSP Field↗

DOE OSTI2025

Scaling deep learning for material imaging with a pseudo 3D model for domain transfer

The recent introduction of deep learning methods for image processing has greatly advanced the characterization of materials using three-dimensional (3D) X-ray imaging techniques. However, deep learning models often have difficulty performing consistently across images owing to unavoidable variations in imaging conditions, which create inconsistencies even for the same material. As a result, networks must frequently be retrained for new datasets, limiting their applicability and generalization. Thus, it is critical to reduce the variations between images to enable a single model to process multiple datasets. Herein, we introduce P3T-Net, a pseudo-3D domain transfer network that transfers diverse 3D images into a uniform domain before processing using deep learning models. Remarkably, P3T-Net enables the reuse of previously trained networks for processing new images and considerably reduces the computational cost of transferring 3D images across domains. These unique capabilities were demonstrated in the following scenarios: (i) image enhancement of fast scans for geological rock and hydrogen fuel cells, (ii) enhancement of images to match the quality of multi-source imaging for lithium-ion batteries, (iii) accurate segmentation of images captured under different conditions, and (iv) tera-scale 3D transfer (10 11 voxels) on a single GPU. Overall, the proposed approach addresses cross-domain inconsistencies across various materials and conditions, thereby enabling more robust and generalizable deep learning solutions for a wide range of material imaging tasks.

25 ENERGY STORAGE↗

DOE OSTI2026

Associated Particle Imaging of Neutron Inelastic Scatter: 3-D Reconstruction, Capabilities, and Challenges

Associated particle imaging (API) offers unique advantages for 3-D imaging of neutron inelastic scatter, including single-view tomographic imaging and data acquisition when access is limited to only one side of the interrogated object. However, widespread adoption of neutron inelastic scatter imaging has been impeded by several inherent challenges, most prominently spatial resolution, self-attenuation, and statistical noise. Here, in this work, the capabilities and challenges of neutron inelastic scatter imaging are investigated. Instead of focusing on a single imaging application, we identify imaging principles that hold for various neutron inelastic scatter imaging techniques. The primary challenges for 3-D imaging are characterized. The inherent spatial resolution in the time-of-flight (TOF) dimension is derived based on the known system timing resolution and scan geometry. Three reconstruction algorithms are described and demonstrated, including the introduction of modern iterative reconstruction incorporating a physics-based system model. Simulation is leveraged to demonstrate imaging capability with varying coincidence count levels and system timing resolution. An example of measured data with both back-scatter and forward-scatter detector positioning is presented. System design characteristics and their effects on image quality are identified. The imaging framework presented in this article has the potential to facilitate growth of 3-D neutron inelastic scatter API by identifying applications that are a good match for the technique and by targeting system development resources toward the requirements of a specific imaging task.

Associated particle imaging (API)↗

DOE OSTI2026

Ray-tracing image simulations of transparent objects with complex shape and inhomogeneous refractive index

Optical images of transparent three-dimensional objects can be different from a replica of the object’s cross section in the image plane, due to refraction at the surface or in the body of the object. Simulations of the object’s image are thus needed for the visualization and validation of physical models. We report ray tracing image simulations that achieved high physical fidelity, reproducing optical behaviors and image features not rendered in previous studies. We replicated brightfield microscopy images of drops with complex shapes, and images of pressure and shock waves traveling inside them. For high physical fidelity, the simulations must replicate the spatial and angular distribution of illumination rays, and both the experiment and the simulation must be designed for accurate optical modeling. The simulations are highly sensitive to the properties of the drops and can be used to diagnose and refine fluid dynamics models. The simulated images can also be optimized to extract multiple 3D properties from experimental images. Compared to specialized single-shot 3D imaging methods, this approach has the advantage that it preserves the experimental simplicity, the high resolution, and the visual interpretability characteristic to basic optical imaging. The techniques introduced here are directly applicable to optical microscopy, so they can be used in other fields, such as microfluidics and biology, to expand the type and the accuracy of three-dimensional information that can be extracted from basic optical images.

Cavitation↗

DOE OSTI2026

A unified large language model–based framework for heterogeneous PV image diagnosis

With advances in imaging technologies, modern photovoltaic (PV) systems generate large volumes of heterogeneous image data, including visible, electroluminescence (EL), and infrared (IR) images. Existing PV image analysis models, particularly deep learning approaches, are typically task-specific and lack cross-modality generalization. To address this limitation, this paper proposes an open-source large language model (LLM)–based unified framework for heterogeneous PV image diagnostics. Through task-aware diagnostic prompting, the framework enables analysis of visible, EL, and IR images within a single pipeline, supporting both zero-shot and few-shot inference and binary and multiclass classification. It is compatible with state-of-the-art multimodal LLMs, including ChatGPT, Gemini, Claude, Qwen, and CLIP. The framework is evaluated on PV module condition classification (clean, soiling, snow, hail, and bird droppings) using visible images, cell crack detection using EL images, and hotspot detection using IR images. GPT-5.1 in few-shot mode achieves the best performance, with classification accuracy exceeding 97.3%. Open-source models such as Qwen and CLIP also deliver competitive results on visible images (around 90% accuracy), though their performance is more limited on EL and IR modalities. On the full ELPV dataset, the framework achieves 83.5% zero-shot accuracy, within 2.8% of the supervised CNN baseline, confirming scalability to larger benchmarks. Practical aspects such as reproducibility, response latency, and confidence estimation are systematically analyzed. The framework operates across PV image modalities without modality- or task-specific training, making it well suited as a rapid pre-screening tool to support downstream detailed diagnostics. A benchmark dataset of diverse labeled PV images is also released.

Li, Baojie↗

DOE OSTI2025

Mask-side Hyper-NA EUV imaging on the SHARP microscope

Hyper-NA, the prospective successor to high-numerical aperture (NA) extreme ultraviolet lithography (EUVL) could be inserted soon after 2030. Hyper-NA poses a number of challenges, including reduced depth of focus, amplified mask three-dimensional effects, and increased mask-side angular range. A Hyper-NA capable extreme ultraviolet (EUV) mask-imaging tool can address these challenges and accelerate research and development toward Hyper-NA. The Sharp High-Numerical Aperture Actinic Reticle Review Project (SHARP) EUV mask microscope is supporting mask-side high-NA imaging since 2015. Implementing mask-side Hyper-NA imaging in 2024 enables research and development toward the corresponding nodes of EUVL. Hyper-NA zoneplates at 0.75 4x/8x NA with a 6.7-deg chief ray angle and 0.85 4x/8x NA with a 7.4-deg chief ray angle are added to the SHARP microscope. Imaging at mask-side Hyper-NA is demonstrated. Imaging of 5-nm half-pitch (wafer scale) horizontal lines and spaces is demonstrated using dipole illumination. Imaging of 5-nm half-pitch (wafer scale) vertical lines and spaces is demonstrated using frequency-doubled imaging of 40-nm hp (mask scale) lines and spaces. Normalized image log slope (NILS) and modulation of Hyper-NA image data match closely to simulations for horizontal lines and spaces. A reduction in NILS of 0.3 or less is observed for vertical lines and spaces in the two-beam imaging regime. Through-focus image data are discussed, comparing different dipole sources and mask-side NAs. Mask-side Hyper-NA photomask imaging has been implemented and demonstrated on the SHARP microscope and is now available to users of the instrument.

Benk, Markus↗

DOE OSTI2025

Time-resolved detectors for quantum ghost imaging

Quantum ghost imaging is a method that utilizes the correlated detection of two photons to generate an image. One photon is detected by an imaging sensor and the other by a single-element bucket detector. The selection of the imaging sensor and its capabilities relative to the bucket detector impact the quality of the ghost images. This work examines a SPAD array and a photocathode detector as imaging sensors for quantum ghost imaging. We discuss how to achieve optimal images using these two technologies. We also demonstrate that these devices are able to generate ghost images at 1Hz frame rates, expanding the technique to biologically relevant time scales.

36 MATERIALS SCIENCE↗

DOE OSTI2026

INR-TEM: Robust cavity detection in multifocus TEM images via implicit neural representations

When characterizing materials using transmission electron microscopy (TEM) images, detecting and quantifying small features in microstructures, such as cavities, pose significant challenges. Off-the-shelf object detection models, including YOLOv8, show considerable performance degradation, particularly when images vary in resolution and the objects of interest possess a low percentage of the total image region of interest. In this study, we introduce a novel detection pipeline that incorporates an implicit neural representation (INR)-based detection method, INR-TEM, and two-modality imaging (e.g., under-focused and over-focused images typically acquired during materials characterization) to improve object detection performance. The INR-TEM method incorporates a pixel-wise prediction principle inspired by pixel-wise centerness weighting. INR-TEM demonstrates superior robustness to resolution variability, maintaining high detection accuracy even at low image resolutions compared to YOLOv8. To leverage INR-TEM effectively in real-world two-modality characterization applications, we further integrate a two-stage motion correction pipeline designed explicitly for aligning multifocus TEM images. The alignment process, comprising keypoint (based on scale-invariant feature transform, SIFT) and intensity matching, significantly mitigates the adverse effects of perceived motion-induced image degradation during through-focal TEM imaging, directly enhancing INR-TEM’s detection capability over conventional single-focus images. Our integrated INR-TEM cavity detection framework notably improves performance across various cavity sizes, outperforming off-the-shelf YOLOv8 detections that rely on a single image modality.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

DOE OSTI2024

Report on the AAPM grand challenge on deep generative modeling for learning medical image statistics

Abstract Background The findings of the 2023 AAPM Grand Challenge on Deep Generative Modeling for Learning Medical Image Statistics are reported in this Special Report. Purpose The goal of this challenge was to promote the development of deep generative models for medical imaging and to emphasize the need for their domain‐relevant assessments via the analysis of relevant image statistics. Methods As part of this Grand Challenge, a common training dataset and an evaluation procedure was developed for benchmarking deep generative models for medical image synthesis. To create the training dataset, an established 3D virtual breast phantom was adapted. The resulting dataset comprised about 108 000 images of size 512 512. For the evaluation of submissions to the Challenge, an ensemble of 10 000 DGM‐generated images from each submission was employed. The evaluation procedure consisted of two stages. In the first stage, a preliminary check for memorization and image quality (via the Fréchet Inception Distance [FID]) was performed. Submissions that passed the first stage were then evaluated for the reproducibility of image statistics corresponding to several feature families including texture, morphology, image moments, fractal statistics, and skeleton statistics. A summary measure in this feature space was employed to rank the submissions. Additional analyses of submissions was performed to assess DGM performance specific to individual feature families, the four classes in the training data, and also to identify various artifacts. Results Fifty‐eight submissions from 12 unique users were received for this Challenge. Out of these 12 submissions, 9 submissions passed the first stage of evaluation and were eligible for ranking. The top‐ranked submission employed a conditional latent diffusion model, whereas the joint runners‐up employed a generative adversarial network, followed by another network for image superresolution. In general, we observed that the overall ranking of the top 9 submissions according to our evaluation method (i) did not match the FID‐based ranking, and (ii) differed with respect to individual feature families. Another important finding from our additional analyses was that different DGMs demonstrated similar kinds of artifacts. Conclusions This Grand Challenge highlighted the need for domain‐specific evaluation to further DGM design as well as deployment. It also demonstrated that the specification of a DGM may differ depending on its intended use.

Radiology, Nuclear Medicine & Medical Imaging↗

DOE OSTI2025

Image registration for accurate electrode deformation analysis in operando microscopy of battery materials

Operando imaging techniques have become increasingly valuable in both battery research and manufacturing. However, the reliability of these methods can be compromised by instabilities in the imaging setup and operando cells, particularly when utilizing high-resolution imaging systems. The acquired imaging data often include features arising from both undesirable system vibrations and drift, as well as the scientifically relevant deformations occurring in the battery sample during cell operation. For meaningful analysis, it is crucial to distinguish and separately evaluate these two factors. To address these challenges, we employ a suite of advanced image-processing techniques. These include fast Fourier transform analysis in the frequency domain, power spectrum-based assessments for image quality, as well as rigid and non-rigid image-registration methods. These techniques allow us to identify and exclude blurred images, correct for displacements caused by motor vibrations and sample holder drift and, thus, prevent unwanted image artifacts from affecting subsequent analyses and interpretations. Additionally, we apply optical flow analysis to track the dynamic deformation of battery electrode materials during electrochemical cycling. This enables us to observe and quantify the evolving mechanical responses of the electrodes, offering deeper insights into battery degradation. Together, these methods ensure more accurate image analysis and enhance our understanding of the chemomechanical interplay in battery performance and longevity.

Sun, Tianxiao↗

DOE OSTI2025

Advancing X-ray quantum imaging through Monte-Carlo simulations

Imaging with X-rays poses fundamental limits due to radiation damage of the highly energetic photons. This becomes problematic for sensitive biological systems such as subcellular structures. Lowering the radiation dose, without sacrificing the signal-to-noise ratio, would be desirable for any kind of imaging modalities involving X-rays. To achieve this goal, quantum imaging with entangled X-ray photons constitutes a promising route. Production of biphotons have been demonstrated in the X-ray regime by the process of Spontaneous Parametric Down-Conversion (SPDC). However, compared to SPDC in the regime of visible light, the production rate for X-ray biphotons is extremely low. With the introduction of new high average brightness X-ray sources, such as 4th generation synchrotrons and high repetition rate Free-Electron X-ray Lasers (XFEL), quantum imaging may become practical. We introduce a ray tracing approach using Monte-Carlo sampling, specifically designed for quantum imaging with entangled X-ray photons generated by SPDC. By simulation, the superior image quality of quantum over classical imaging methods is demonstrated using realistic experimental conditions available at high repetition rate XFELs. With these simulations, we can efficiently assist the design of future experiments at beam lines, which can substantially accelerate the advancement of X-ray quantum imaging and reduce costs.

Entangled Photons↗

DOE OSTI2025

Fresnel diffraction imaging of surface nanostructure using coherent resonant x-ray scattering

We investigated surface nanostructures on an antiferromagnet MnBi 2 Te 4 using a novel imaging technique, direct (real)-space and real time coherent x-ray imaging (direct-CXI). This technique has provided new insights into antiferromagnetic textures, including the formation of anti-phase antiferromagnetic (AFM) domains and thermal dynamics of AFM domains and domain walls. While this method produces real-space images of AFM textures without requiring a complex imaging retrieval process, its underlying imaging mechanism has not been fully understood, limiting a deep understanding of AFM textures and the information they contain. By investigating the well-defined structural characteristics of the nanostructures fabricated on MnBi 2 Te 4 ⁠, we elucidate the imaging principle of this novel technique. We find that the observed images can be well explained by the Fresnel diffraction integral. Using a simple model from classical optics, our calculations successfully reproduce the experimentally observed images of the nanostructures. This demonstrates that direct-CXI not only provides straightforward real-space imaging but also contains phase information through its Fresnel diffraction integral.

36 MATERIALS SCIENCE↗

DOE OSTI2025

Automated Scoring of Morphological Changes in Images of Pentaerythritol Tetranitrate

Recent advances in characterization techniques that generate large datasets of material microstructure images require robust, automated image-processing. We applied an unsupervised anomaly detection method called feature anomaly detection system (FADS) to automatically detect and quantify microstructure changes in images of the explosive pentaerythritol tetranitrate (PETN) aged at various temperatures. We demonstrated the FADS approach on two-dimensional images extracted from computed tomography scans, but the same technique can be readily applied to other imaging modalities. FADS calculates anomaly scores on the basis of differences in filter activations of nominal and test data in pretrained convolutional neural networks. The FADS scores successfully differentiated between pristine PETN and PETN aged at a temperature where material coarsening occurred. Morphological metric analysis of segmented images verified observed trends in FADS scores as a function of aging temperature and aging time, specifically by calculating volume fractions, specific boundary lengths, two-point correlation functions, and local thicknesses. Here, the FADS technique has two important advantages compared to traditional morphological analysis: First, it uses grayscale images as input, rather than images that are segmented to separate the appropriate phases; and second, FADS scores capture any type of changes among image sets, rather than requiring prior knowledge or selection of a relevant set of metrics.

Accelerated aging↗

DOE OSTI2025

Forward modeling fluctuations in the DESI LRGs target sample using image simulations

We use the forward modeling pipeline, Obiwan, to study the imaging systematics of the Luminous Red Galaxies (LRGs) targeted by the Dark Energy Spectroscopic Instrument (DESI). Imaging systematics refers to the false fluctuation of galaxy densities due to varying observing conditions and astrophysical foregrounds corresponding to the imaging surveys from which DESI LRG target galaxies are selected. We update the Obiwan pipeline, which we previously developed to simulate the optical images used to target DESI data, to further simulate WISE images in the infrared. This addition allows simulating the DESI LRGs sample, which utilizes WISE data in the target selection. Deep DESI imaging data combined with a method to account for biases in their shapes is used to define a truth sample of potential LRG targets. We inject these data evenly throughout the DESI Legacy Imaging Survey footprint at declinations between -30 and 32.375 degrees. We simulate a total of 15 million galaxies to obtain a simulated LRG sample (Obiwan LRGs) that predicts the variations in target density due to imaging properties. We find that the simulations predict the trends with depth observed in the data, including how they depend on the intrinsic brightness of the galaxies. We observe that faint LRGs are the main contributing source of the imaging systematics trend induced by depth. We also find significant trends in the data against Galactic extinction that are not predicted by Obiwan. These trends depend strongly on the particular map of Galactic extinction chosen to test against, implying systematic contamination in the Galactic extinction maps is a likely root cause (e.g., Cosmic-Infrared Background, dust temperature correction). We additionally observe a morphological change of the DESI LRGs population evidenced by a correlation between OII emission line average intensity and the size of the z-band PSF. This effect most likely results from uncertainties in background subtraction. The detailed findings we present should be used to guide any observational systematics mitigation treatment for the clustering of the DESI LRGs sample.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2025

Neutron and Gamma-Ray Imaging of Th-232 and Cm-244 Using Organic Glass Scintillators

Modern nuclear safeguards require detection and characterization capabilities suitable for a wide variety of radiation sources and applications. Field-deployable detection systems have also had to modernize to meet changing needs. Recent developments in organic scintillator technology have resulted in the creation of an organic glass scintillator (OGS) at Sandia National Laboratory which is composed of a 9:1 mixture of glass compounds C42H36Si and C51H44Si. The novel scintillator composition was implemented into the design for a dual-particle capable imaging system at the University of Michigan. This work presents new results from two experiments demonstrating the gamma-ray and fast-neutron imaging capabilities of the organic glass system. Gamma spectroscopy was also performed using CeBr3 scintillators that are part of the imager design. Measurements were performed at Lawrence Livermore National Laboratory using 232Th metal hemishells and an encapsulated 244Cm oxide source. Successful gamma-ray imaging of the 232Th distributed sources is demonstrated with the glass imager, but there were no appreciable neutrons from the 232Th for neutron imaging. Promising neutron and gamma-ray imaging results of 244Cm are demonstrated despite limited imaging event statistics available in this measurement. Gamma-ray spectroscopy results were able to identify 232Th using prominent emissions at 239, 338, 583, and 911 keV. 244Cm was identified from emissions of 43, 99, and 153 keV. These results demonstrate the potential of organic glass imaging for nuclear nonproliferation or characterization efforts.

Lopez, Ricardo [Univ. of Michigan, Ann Arbor, MI (↗

DOE OSTI2025

Ultra-high Information-content Chemical Imaging with Broadband Coherent Anti-Stokes Raman and Two-photon Fluorescence Lifetime Microscopy

Raman fingerprint spectroscopy and fluorescence lifetime imaging are emerging tools for studying metabolic profiles of biological specimens. While Raman fingerprint spectroscopy detects intrinsic molecular vibrations that reflect the molecular composition and chemical environment of a sample, fluorescence lifetime imaging measures changes in the excited-state lifetime of fluorophores that are sensitive to their microenvironments. Here, we present a multimodal imaging platform combining broadband coherent anti-Stokes Raman scattering (BCARS) and two-photon fluorescence lifetime imaging (2p-FLIM) microscopy that can acquire biologically relevant Raman fingerprint spectra and fluorescence lifetime signals in vivo and simultaneously. The tremendous chemical information obtained from spatially co-registered BCARS and 2p-FLIM images allows us to characterize the subtle differences between sub-cellular compartments and verify the potential false-positive results generated by fluorescence imaging alone. This is demonstrated by directly comparing the BCARS, 2p-FLIM, and two-photon excitation fluorescence(TPEF) signals simultaneously obtained from the same dye-stained organelle in the live, intact C. elegans expressing a green fluorescent protein (GFP) marker. In this work, we introduce the BCARS/2p-FLIM/TPEF setup scheme, the image acquisition steps, data processing, and representative results showing that the cross-modality imaging method enables rigorous characterization and in vivo detection at sub-cellular resolution. Furthermore, this protocol provides a framework for simultaneous chemical and fluorescence lifetime imaging to improve the accuracy of biological interpretation in complex living systems.

Xu, Haoyu [Georgia Institute of Technology, Atlant↗

DOE OSTI2026

Original images, ground truth annotations, and precited masks from U-NET for a fungal species x nitrogen experiment

This document describes the “MicroVision-MV003” dataset. This folder contains three subfolders: “originals”, “predicted_masks”, and “gt_masks”. The “originals” folder contains 720 original images acquired using two imaging modalities: 360 images from overhead imaging and 360 images from transmission (raw) imaging. Naming scheme:YYYY-MM-dd__plate_strain_nitrogen-level_replication_experiment_imaging-modalityYYYY-MM-dd: 2024-12-20, 2024-12-21, 2024-12-22,2024-12-23,2024-12-24,2024-12-25,2024-12-26,2024-12-27,2024-12-28,2024-12-29,2024-12-30,2024-12-31.plate: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30.strain: FG, LE.nitrogen-level: N-1, N-10, N-100.replication: 1, 2, 3, 4, 5.experiment: MV.003.imaging-modality: overhead, transmission (raw). For example, the image2024-12-20__1_FG_N-10_3_MV.003_overheadwas taken on 2024-12-20. The plate number is 1, the fungal strain is FG, the nitrogen treatment is N-10, and the replication number is 3 for experiment MV.003, acquired using the overhead imaging modality. The “predicted_masks” folder contains masks generated by a trained U-Net model for the transmission imaging modality.The “gt_masks” folder contains 326 human hand-traced masks that serve as ground truth.

09 BIOMASS FUELS↗

DOE OSTI2024

Uncertainty-refined image segmentation under domain shift

Digital image segmentation is provided. The method comprises training a neural network for image segmentation with a labeled training dataset from a first domain, wherein a subset of nodes in the neural net are dropped out during training. The neural network receives image data from a second, different domain. A vector of N values that sum to 1 is calculated for each image element, wherein each value represents an image segmentation class. A label is assigned to each image element according to the class with the highest value in the vector. Multiple inferences are performed with active dropout layers for each image element, and an uncertainty value is generated for each image element. Uncertainty is resolved according to expected characteristics. The label of any image element with an uncertainty above a threshold is replaced with a new label corresponding to a segmentation class based on domain knowledge.

Martinez, Carianne↗