<p>  </p> <p>Semantic segmentation is a machine learning task that is seeing increased utilization in multiples fields, from medical imagery, to land demarcation, and autonomous vehicles. Semantic segmentation performs the pixel-wise classification of images, creating a new, segmented representation of the input that can be useful for detected various terrain and objects within and image. Recently, convolutional neural networks have been heavily utilized when creating neural networks tackling the semantic segmentation task. This is particularly true in the field of autonomous driving systems.</p> <p>The requirements of automated driver assistance systems (ADAS) drive semantic segmentation models targeted for deployment on ADAS to be lightweight while maintaining accuracy. A commonly used method to increase accuracy in the autonomous vehicle field is to fuse multiple sensory modalities. This research focuses on leveraging the fusion of long wave infrared (LWIR) imagery with visual spectrum imagery to fill in the inherent performance gaps when using visual imagery alone. This comes with a host of benefits, such as increase performance in various lighting conditions and adverse environmental conditions. Utilizing this fusion technique is an effective method of increasing the accuracy of a semantic segmentation model. Being a lightweight architecture is key for successful deployment on ADAS, as these systems often have resource constraints and need to operate in real-time. Multi-Spectral Fusion Network (MFNet) [ 1 ] accomplishes these parameters by leveraging a sensory fusion approach, and as such was selected as the baseline architecture for this research.</p> <p>Many improvements were made upon the baseline architecture by leveraging a variety of techniques. Such improvements include the proposal of a novel loss function categorical cross-entropy dice loss, introduction of squeeze and excitation (SE) blocks, addition of pyramid pooling, a new fusion technique, and drop input data augmentation. These improvements culminated in the creation of the Fast Thermal Fusion Network (FTFNet). Further improvements were made by introducing depthwise separable convolutional layers leading to lightweight FTFNet variants, FTFNet Lite 1 & 2.</p>

The large cache of scrolls from Herculaneum were opened to scholars in spite of and because of the destruction of the volcano and damaging unfurling techniques. The fragments inherited have been studied closely by scholars. Digitization and technological work on the Herculaneum papyri, including the important infrared imaging completed by BYU in the early 2000s, and the 3-D imaging and inchoate virtual unwrapping technology by EduceLab, have amplified and aided scholarship on the scrolls and will continue to do so. P.Herc. 817 is a unique Latin text within the collection that has been heavily studied by scholars due to its fascinating subject matter on the Battle of Actium and Cleopatra and its readability. This fragment serves as a case study to demonstrate the value of each set of digital images in transcription and interpretation research, which suggests at the value of the varying digital images for other Herculaneum fragments. I closely compare digital surrogates of P.Herc. 817 including the early 2000s infrared images, 1960's era negatives, and recent 3-D images with the original artifacts as seen at the Biblioteca Nazionale in July 2023. This autopsy of versions of P.Herc. 817 substantiates the need for scholars to use all available digital images in concert with the original papyri when doing scholarly work. It also reinforces the need for digital stewardship and preservation of each distinct image set. Finally, a hypothetical case study is offered to show the loss to scholarship if the digital images and originals were lost and solely secondary sources remained. Each image set offers value and captures a moment in time of the papyri. As technology continues to progress and excitingly unlocks unseen papyri, care needs to be taken to safeguard and digitally preserve the new along with the older data sets.

No description available.

Inledning:Oral cancer är ett allvarligt tillstånd med hög dödlighet, särskilt vid sen diagnostisering. Brasilien är ett av de länder i världen som har högst prevalens och dödlighet i oral cancer och det är den femte vanligaste cancerformen I landet. Ett hjälpmedel för tidig diagnostisering är önskvärd.Syfte:Att utvärdera skillnaden i diagnostisk tillförlitlighet mellan konventionell oral undersökning och användning av multispektralt ljus (Identafi®) som en metod för tidig upptäckt av potentiellt maligna och maligna lesioner i munslemhinnan i en brasiliansk befolkning.Material och metod:Screening av en befolkning med förhöjd risk för att utveckla oral cancer i Goiânia, Goiás, Brasilien, för att upptäcka potentiellt maligna (PML) eller maligna lesioner (ML). Patienter med misstänkta PML eller ML upptäckta under screeningen undersöktes med multispektralt ljus (Identafi®). Tre oberoende observatörer genomförde bedömning med Identafi® och slutgiltig beslut avseende PML/ML togs i konsensus. Biopsier användes som diagnostisk referensstandard. Interobservatörs överensstämmelse beräknades som procentuell överensstämmelse och kappa-värde (κ).Resultat:Undersökning med Identafi® genererade tolv biopsier. Resultaten blev tre sant positiva, fem falskt positiva, två sant negativa och noll falskt negativa. Sensitiviteten beräknades till 0,29. Specificiteten var inte möjlig att beräkna, då det inte fanns några falska negativa resultat. Interobservatörs överensstämmelse för par av observatörer varierade mellan 78-86% och κ-värden mellan 0,46-0,60.Slutsats:Slutsatsen är att multispektralt ljus, Identafi® inte har inga fördelar jämfört med konventionell klinisk undersökning i fråga om diagnostisk träffsäkerhet för PML eller ML. Dock kan det vara till hjälp för en tandläkare eller oral kirurg som stöd i sitt beslutsfattande. Det kan också hjälpa kirurgen att ta en biopsi från det mest misstänkta delen av lesionen. Det finns inte tillräckligt publicerade studier som tyder på att Identafi® kan skilja mellan normal slemhinna och PML eller ML och denna studie bekräftar detta. Användning av Identafi® som ett hjälpmedel vid screening och undersökning för PML eller ML behöver utredas ytterligare, men baserat på denna studie kan Identafi® inte rekommenderas. / Introduction: Oral cancer is a severe condition with high mortality rate, in particular if diagnosed late. Brazil is one of the countries in the world with high prevalence and mortality from oral cancer and it is the fifth most common cancer there. An aid in early detection of oral cancer as an adjunct to health promotion is desirable.Purpose:The aim of this study was to evaluate the diagnostic accuracy of conventional oral examination and the use of multi spectral light (Identafi®) as an approach for early detection of potentially malignant or malignant lesions in the oral mucosa in a Brazilian population.Material and method:Screening of high-risk population in Goiania, Goias, Brazil, for oral potentially malignant lesions (PML) or malignant lesions (ML) as a selection phase. Patients collected from the screening were examined with multi spectral light (Identafi®) to evaluate diagnostic accuracy. Three observers independently assessed all lesions with Identafi® and the final decision if a lesion was present was taken in consensus. Inter observer agreement was calculated as overall agreement and as kappa value (κ). Biopsies were used as diagnostic reference standard.Results:Identafi® generated, from twelve biopsies, three true positive, five false positive, two true negative and zero false negative. Sensitivity was calculated to 0.29 and specificity was not possible to calculate since there were no false negative results.Inter observer agreement for the use of Identafi® was calculated as overall agreement and as kappa value (κ). The overall agreement for the three pairs of observers varied between 78-86% and κ-values between 0.46 and 0.60.Conclusion:The conclusion of this study is that Identafi® does not have any benefits over conventional oral examination in diagnostic accuracy for potentially malignant or malignant lesions in the oral mucosa. It might, however, be an aid for a dentists or oral surgeons that are unsure whether to take a biopsy or not. It can also aid the surgeon when taking a biopsy to take the most suspicious part of the lesion. There are not enough published evidence that Identafi® can discriminate between normal mucosa and PML or ML, and this study confirms previous results. The use of Identafi® as an aid in screening and examination for PML or ML needs further investigation.

The efficient and timely monitoring of forest dynamics is of paramount importance and requires accurate, high-resolution and time-tagged predictions at global scale. Despite numerous methodologies have been proposed in the literature, existing approaches often compromise on accuracy, resolution, temporal fidelity or coverage. To tackle these challenges and limitations, the main objective of this doctoral thesis is the investigation of the potential of artificial intelligence (AI) for the regression of bio-physical forest parameters from spaceborne Earth Observation (EO) data. This work explores for the first time the combined use of TanDEM-X single-pass interferometric products and convolutional neural networks for canopy height estimation at country scale. To achieve this, a novel deep learning framework is proposed, leveraging the capability of deep neural networks to effectively capture the complex spatial relationships between forest properties and satellite data, as well as ensuring the adaptability to different environmental conditions. The design and the understanding of the model is driven by explainable AI principles and by considerations on large-scale forest dynamics, with a great emphasis set on the challenges related to the variable acquisition geometry of the TanDEM-X mission, and by relying on the use of LVIS-derived LiDAR measurements as reference data. Moreover, several investigations are conducted on the adaptability of the developed framework for transferring knowledge to related domains, such as digital terrain model regression and above-ground biomass density estimation. Finally, the capability of the proposed approach to be extended to the use of other EO sensors is also evaluated, with a particular emphasis on the ESA Sentinel-1 and Sentinel-2 missions. The developed deep learning framework sets a solid groundwork for the generation of large-scale products of bio-physical forest parameters from spaceborne EO data. The approach achieves cutting-edge performance, significantly advancing the current state of forest assessment and monitoring technologies.

En Viticulture de Précision, l’imagerie multi-spectrale est principalement utilisée pour des dispositifs de télédétection. Ce manuscrit s’intéresse à son utilisation en proxidétection, pour la caractérisation du feuillage. Il présente un dispositif expérimental terrestre mobile composé d’un GPS, d’une caméra multi-spectrale acquérant des images visible et proche infrarouge, et d’un Greenseeker RT-100 mesurant l’indice Normalized Difference Vegetation Index (NDVI). Ce système observe le feuillage de la vigne dans le plan de palissage, en lumière naturelle. La parcelle étudiée comporte trois cépages (Pinot Noir, Chardonnay et Meunier) plantés en carré latin. En 2013, six jeux de données ont été acquis à différents stades phénologiques.Pour accéder aux propriétés spectrales de la végétation, il est nécessaire de calibrer les images en réflectance. Cela requiert l’utilisation d’une mire de MacBeth comme référence radiométrique. Lorsque la mire est cachée par les feuilles, les paramètres de calibration sont estimés par une interpolation linéaire en fonction des images les plus proches sur lesquelles la mire est visible. La cohérence de la méthode d’estimation employée est vérifiée par une validation croisée (LOOCV).La comparaison du NDVI fournie par le Greenseeker avec celui déterminé via les images corrigées permet de valider les données générées par le dispositif. La polyvalence du système est évaluée via les images où plusieurs indices de végétation sont déterminés. Ils permettant des suivis de croissance de la végétation originaux offrant des potentialités de phénotypage ou une caractérisation de l’état sanitaire de la végétation illustrant la polyvalence et le gain en précision de cette technique. / Mutispectral imaging systems are widely used in remote sensing for Precision Viticulture. In this work, this technique was applied in the proximal sensing context to characterize vine foliage. A mobile terrestrial experimental system is presented, composed of a GPS receiver, a multi-spectral camera acquiring visible and near infrared images, and a Greenseeker RT-100 which measures the Normalized Difference Vegetative Index (NDVI). This optical system observes vine foliage in the trellis plan, in natural sunlight. The experimental field is planted with Chardonnay, Pinot Noir and Meunier cultivars in a latin squared pattern. In 2013, six datasets were acquired at various phenological stages.Spectral properties of the vegetation are accessible on images when they are calibrated in reflectance. This step requires the use of a MacBeth colorchart as a radiometric reference. When the chart is hidden by leaves, the calibration parameters are estimated by simple linear interpolation using the results from resembling images, which have a visible chart. The performance of this method is verified with a cross-validation technique (LOOCV).To validate the data provided by the experimental system, the NDVI given by the Greenseeker was compared to those computed from the calibrated images. The assessment of the versatility of the system is done with the images where several indices were determined. It allows an innovative follow-up of the vegetative growth, and offering phenotyping applications. Moreover, the characterization of the sanitary state of the foliage prove that this technique is versatile and accurate.

In the past decade, camera manufacturers have offered smaller form factors, smaller pixel sizes (leading to higher resolution images), and faster processing chips to increase the performance of consumer cameras. However, these conventional approaches have failed to capitalize on the spatio-temporal redundancy inherent in images, nor have they adequately provided a solution for finding $3$D point correspondences for cameras sampling different bands of the visible spectrum. In this thesis, we pose the following question---given the repetitious nature of image patches, and appropriate camera architectures, can statistical models be used to increase temporal, structural, or spectral resolution? While many techniques have been suggested to tackle individual aspects of this question, the proposed solutions either require prohibitively expensive hardware modifications and/or require overly simplistic assumptions about the geometry of the scene. We propose a two-stage solution to facilitate image reconstruction; 1) design a linear camera system that optically encodes scene information and 2) recover full scene information using prior models learned from statistics of natural images. By leveraging the tendency of small regions to repeat throughout an image or video, we are able to learn prior models from patches pulled from exemplar images. The quality of this approach will be demonstrated for two application domains, using low-speed video cameras for high-speed video acquisition and multi-spectral fusion using an array of cameras. We also investigate a conventional approach for finding 3D correspondence that enables a generalized assorted array of cameras to operate in multiple modalities, including multi-spectral, high dynamic range, and polarization imaging of dynamic scenes.

Ce mémoire est constitué de deux parties relativement indépendantes, dont la première partie porte sur les méthodes de mesure de température par méthodes Multi-Spectrales (pyrométrie optique passive), et la seconde sur la Caractérisation Thermique à haute température par transformations intégrales de matériaux orthotropes. Dans chacune de ces deux parties, les méthodes/modèles développés ont été traités du point de vue théorique, numérique, et expérimental. Dans la partie multi-spectrale, une méthode de mesure de température permettant de prendre en compte les variations spectrales de la chaine de mesure globale (incluant l’émissivité) a été présentée. De plus, une méthode de détermination des longueurs d’ondes optimales au sens de la minimisation de l’écart-type sur la température, a été développée. Enfin, il a également été montré que les longueurs d’ondes optimales pour les mesures mono-spectrales et bi-spectrales pouvaient être déterminées à l’aide de lois analogues à la loi de déplacement de Wien. Dans la partie Caractérisation Thermique, différentes méthodes et modèles ont été développés. Les méthodes proposées effectuent l’estimation des diffusivités longitudinales et transversales sur l’ensemble des harmoniques simultanément. De plus, ces méthodes permettent de s’affranchir du couplage thermique dû à la présence d’un porte-échantillon, et/ou d’effectuer des mesures de diffusivités pseudo-locales, en injectant comme conditions aux limites les informations expérimentales obtenues par caméra infrarouge. Enfin, les notions de corrélation entre les paramètres et de durée d’exploitabilité des harmoniques ont également été abordées / This thesis consists of two relatively independent parts, the first part focuses on methods of temperature measurement using Multi-Spectral (passive optical pyrometry) methods, and the second on the Thermal Characterization by integral transforms at high temperature of orthotropic materials. In each of these two parts, methods / models developed were treated from a theoretical point of view, numerical and experimental. In the multi-spectral part, a method of temperature measurement to take into account a spectral variation of the overall measurement chain (including the emissivity) was introduced. Moreover, a method of determining the optimal wavelengths in the sense of minimizing the standard deviation of temperature, has been developed. Finally, it has also been shown that the optimal wavelengths for mono-spectral and bi-spectral measurements could be determined with similar laws to Wien's displacement law. In the Thermal Characterization part, different methods and models have been developed. The proposed methods perform the estimation of longitudinal and transverse diffusivities on all harmonics simultaneously. Furthermore, they allow overcoming the thermal coupling due to the presence of a sample holder, and / or making pseudo-local measurements of diffusivities. Finally, the concepts of correlation between parameters and duration of harmonics exploitability were also discussed.This thesis consists of two relatively independent parts, the first part focuses on methods of temperature measurement using Multi-Spectral (passive optical pyrometry) methods, and the second on the Thermal Characterization by integral transforms at high temperature of orthotropic materials. In each of these two parts, methods / models developed were treated from a theoretical point of view, numerical and experimental. In the multi-spectral part, a method of temperature measurement to take into account a spectral variation of the overall measurement chain (including the emissivity) was introduced. Moreover, a method of determining the optimal wavelengths in the sense of minimizing the standard deviation of temperature, has been developed. Finally, it has also been shown that the optimal wavelengths for mono-spectral and bi-spectral measurements could be determined with similar laws to Wien's displacement law. In the Thermal Characterization part, different methods and models have been developed. The proposed methods perform the estimation of longitudinal and transverse diffusivities on all harmonics simultaneously. Furthermore, they allow overcoming the thermal coupling due to the presence of a sample holder, and / or making pseudo-local measurements of diffusivities. Finally, the concepts of correlation between parameters and duration of harmonics exploitability were also discussed

No description available.

In this work, multi-band (multi-color) detector structures considering different semiconductor device concepts and architectures are presented. Results on detectors operating in ultraviolet-to-infrared regions (UV-to-IR) are discussed. Multi-band detectors are based on quantum dot (QD) structures; which include quantum-dots-in-a-well (DWELL), tunneling quantum dot infrared photodetectors (T-QDIPs), and bi-layer quantum dot infrared photodetectors (Bi-QDIPs); and homo-/heterojunction interfacial workfunction internal photoemission (HIWIP/HEIWIP) structures. QD-based detectors show multi-color characteristics in mid- and far-infrared (MIR/FIR) regions, where as HIWIP/HEIWIP detectors show responses in UV or near-infrared (NIR) regions, and MIR-to-FIR regions. In DWELL structures, InAs QDs are placed in an InGaAs/GaAs quantum well (QW) to introduce photon induced electronic transitions from energy states in the QD to that in QW, leading to multi-color response peaks. One of the DWELL detectors shows response peaks at ∼ 6.25, ∼ 10.5 and ∼ 23.3 µm. In T-QDIP structures, photoexcited carriers are selectively collected from InGaAs QDs through resonant tunneling, while the dark current is blocked using AlGaAs/InGaAsAlGaAs/ blocking barriers placed in the structure. A two-color T-QDIP with photoresponse peaks at 6 and 17 µm operating at room temperature and a 6 THz detector operating at 150 K are presented. Bi-QDIPs consist of two layers of InAs QDs with different QD sizes. The detector exhibits three distinct peaks at 5.6, 8.0, and 23.0 µm. A typical HIWIP/HEIWIP detector structure consists of a single (or series of) doped emitter(s) and undoped barrier(s), which are placed between two highly doped contact layers. The dual-band response arises from interband transitions of carriers in the undoped barrier and intraband transitions in the doped emitter. Two HIWIP detectors, p-GaAs/GaAs and p-Si/Si, showing interband responses with wavelength thresholds at 0.82 and 1.05 µm, and intraband responses with zero response thresholds at 70 and 32 µm, respectively, are presented. HEIWIP detectors based on n-GaN/AlGaN show an interband response in the UV region and intraband response in the 2-14 µm region. A GaN/AlGaN detector structure consisting of three electrical contacts for separate UV and IR active regions is proposed for simultaneous measurements of the two components of the photocurrent generated by UV and IR radiation.