Дисертації з теми "Task-condition"

Understanding road surface conditions is an important component in active vehicle safety. Estimations can be achieved through image classification using increasingly popular convolutional neural networks (CNNs). In this paper, we explore the effects of multi-task learning by creating CNNs capable of simultaneously performing the two tasks road surface condition classification (RSCC) and road scene semantic segmentation (RSSS). A multi-task network, containing a shared feature extractor (VGG16, ResNet-18, ResNet-101) and two taskspecific network branches, is built and trained using the Road-Conditions and Cityscapes datasets. We reveal that utilizing task-dependent homoscedastic uncertainty in the learning process improvesmulti-task model performance on both tasks. When performing task adaptation, using a small set of additional data labeled with semantic information, we gain considerable RSCC improvements on complex models. Furthermore, we demonstrate increased model generalizability in multi-task models, with up to 12% higher F1-score compared to single-task models.

The aim of the study was to investigate the association between working memory capacity and golf performance in a dual-task paradigm. Twenty-eight golfers with ages varying from 19-58 participated in the study. The participants were instructed to, after assessing their working memory capacity with digit span tests, hit ten golf putts from varying distances in two different conditions, one single-task condition (just performing the golf putts) and one dual-task condition (golf putts plus a working memory task). In the dual-task condition the participants were given an address (including street name, postal code and city) to remember while hitting all ten putts. After hitting the putts in each condition the participants reported their perceived mental effort. Results from the study showed that participants with higher working memory capacity performed better than participants with lower working memory capacity in general, but especially in the distracted condition. Results also showed that participants with higher working memory capacity perceived less mental effort compared to participants with lower working memory capacity, especially in the distracted condition. The findings from this study indicate that working memory capacity is related to performance and mental effort exerted during performance, especially in a distracted condition.

This study examined the effects of three fixed-duration free operant access conditions on rates of automatically maintained stereotypic behavior and correct task responding during discrete trial training (DTT) with two children diagnosed with autism. Following a functional analysis, confirming automatic function, interviews/observations were conducted to identify behavioral indicators of satiation and an average satiation level. In this endeavor, participants were exposed to a free operant condition to validate satiation of stereotypic responding. Once satiation level was averaged, two durational conditions were computed: Long (75% access) and Short (25% access). A third condition, Deprivation, involved blocking all attempts at the stereotypic response for the average duration till satiation. An alternating treatment design was used to examine the effects of these three conditions on stereotypy and correct responding during subsequent discrete trial tasks. For both, participants correct responding did not seem to be affected by the length of the pre-session access to the stereotypic behavior prior to the DTT session. For Marcus, the Long condition may have acted as an abolishing operation (AO) during DTT. Following the Short condition he engaged in higher rates of the stereotypic behaviors during his DTT sessions compared to the other conditions. For the Sara, it appears that pre-session access to stereotypy had little effect on stereotypic behavior during DTT sessions.

Mind-wandering was until recently not a mainstream topic of research. The aim of this literature review is to present current views on the definition of mind-wandering and how the phenomenon is experienced. Furthermore, it gives an account of the implications of mind-wandering on cognitive performance, as well as its neurological correlates. In addition, the methods used to study mind-wandering are reviewed.The study of mind-wandering reveals a highly frequent phenomenon with practical consequences on a broad scale, both disruptive and supportive to goal-related behaviour and wellbeing in general. Originating from the default network, and its regions related to representations of self, memory, Theory of Mind, empathy and creativity, mind-wandering is hypothesized to be a function for planning one’s future life. Suggested further research concerns how mind-wandering can be countered, detected from the outside and whether it alters the physical feature of the brain.

De nombreuses études ont montré que l’activité physique régulière et le niveau de condition physique étaient associés à la performance cognitive et plus particulièrement aux fonctions exécutives. Parmi les nombreux mécanismes physiologiques qui sous-tendent cette association, l’oxygénation du cortex préfrontal semble avoir un rôle majeur. Il est bien établi que l'impact spécifique de l'oxygénation préfrontale sur le lien entre l’activité physique et la cognition est influencé par certains facteurs comme le sexe, l’âge, et la santé cardio-vasculaire. Ainsi, l'objectif de cette thèse était d'étudier l'impact de l'activité physique et du niveau de condition physique sur les fonctions exécutives au cours du vieillissement chez des hommes sains. L'influence de l'oxygénation du cortex préfrontal et de la santé cardiovasculaire sur la relation « niveau de condition physique – fonctions exécutives » a également été étudiée. A partir de la littérature existante, nous avons formulé l’hypothèse que l’activité physique et le profil cardiorespiratoire auraient des effets bénéfiques sur les fonctions exécutives chez les hommes jeunes et plus âgés, et que pour ces deux populations, l’oxygénation du cortex préfrontal et la santé cardiovasculaire seraient impliquées dans cette relation. Une revue de littérature et quatre études transversales ont été réalisées pour vérifier ces hypothèses.Notre revue de littérature chez les personnes âgées a indiqué que:• Un niveau de condition physique plus élevé est associé à de meilleures performances dans certains tests de fonction exécutive.• Un niveau de condition physique plus élevé est associé à une moindre rigidité artérielle, à une réactivité vasculaire plus élevée, et à une plus grande amplitude de l'oxygénation cérébrale au cours de l'exercice ou d'une tâche cognitives.• Un plus grand volume de matière grise et une plus grande d’intégrité de la substance blanche sont liés àu profil cardiorespiratoire, mais moins systématiquement associés à l’activité physique.• Au moins de 12 semaines d'un programme d'exercice aérobie sont nécessaires pour obtenir un effet significatif sur les fonctions cognitivesNos études expérimentales nous ont permis de conclure que:Chez les jeunes hommes:• Les individus actifs obtenaient de meilleures performances dans les tâches exécutives que les inactifs et avaient un changement important dans l'oxygénation du cortex préfrontal lors des conditions les plus complexes de la tâche de Stroop.• Un niveau de condition physique plus élevé était liée à une meilleure performance en double tâche et une plus grande oxygénation des deux côtés du cortex préfrontal.Chez les hommes âgés:• Le niveau de condition physique n’est pas lié aux performances de Stroop ni à l’ oxygénation du cortex préfrontal.• Un niveau de condition physique plus élevé est lié à une meilleure performance et une plus grande oxygénation de cortex préfrontal droit au cours d’une tâche de Stroop dans le groupe de 61+ ans mais pas dans le groupe de 55-60 ans, ce qui suggère l'importance de la classification de groupe d'âge pour évaluer l'effet du niveau de condition physique sur les fonctions exécutives chez les hommes âgés.• Les hommes âgés ont des relations plus fortes entre plusieurs paramètres cardiovasculaires et la performance des tâches de Stroop que les jeunes hommesCette projet de recherche montre que l'activité physique et le niveau de condition physique ont des effets positifs sur les fonctions exécutives chez les jeunes hommes et âgés, en particulier dans la tâche la plus complexes. L’oxygénation du cortex préfrontal et la santé cardiovasculaire modulent la relation entre l'activité physique et la cognition. Nous concluons que d’être physiquement actif ou d'avoir un meilleur niveau de condition physique donne des effets avantageux pour la santé vasculaire, l’oxygénation du cortex préfrontal, et les fonctions exécutives non seulement chez les hommes âgés, mais aussi chez les jeunes hommes
Many studies have reported that regular physical activity and cardiorespiratory fitness were associated with cognitive performance and more selectively with executive functions. Among numerous physiological mechanisms that may underlie the association between them, prefrontal cortex oxygenation seems to play a major role. However, the specific impact of prefrontal oxygenation on the link between physical activity and cognition is influenced by several factors, such as gender, age, or cardiovascular health. Therefore, the aim of this thesis was to investigate the impact of physical activity and cardiorespiratory fitness on executive functions across the adults' age span in healthy males, as well as the influence of prefrontal cortex oxygenation and cardiovascular health. To obtain those objectives, we conducted a review of the effect of physical activity on the brain in older adults and four cross-sectional studies. From our review, we highlighted that: • In older adults, higher fitness level is associated with better performance in several executive function tests. Even though the limited number of studies available makes it difficult to draw definitive conclusions.• Better cardiovascular fitness in older adults is associated with improve arterial stiffness, higher vascular reactivity, and greater amplitude of cerebral oxygenation during exercise or cognitive tasks.• Greater gray matter volume and white matter integrity were related to the cardiorespiratory fitness but less consistently related to physical activity.• At least 12 weeks of an aerobic exercise program are required to give advantageous effects to the brainAnd our experimental works show that:In young males:• The active individuals performed better in executive tasks than their inactive counterparts and had a larger change in prefrontal cortex oxygenation during the most complex conditions of Stroop task.• High cardiorespiratory fitness was related to a better performance in dual-task and greater oxygenation on both sides of the prefrontal cortex.In older males:• Cardiorespiratory fitness is not related to Stroop task performance nor prefrontal cortex oxygenation in overall older males.• Higher cardiorespiratory fitness was related to a better performance and greater right prefrontal cortex oxygenation during a Stroop task in 61+ years old group but not in 55-60 years old group, suggesting the importance of age-group classification to evaluate the effect of cardiorespiratory fitness on executive function in older male subjects.• Older males have stronger relationships between several vascular parameters and Stroop task performance than young malesThis work shows the relationship between physical activity and cardiorespiratory fitness on executive functions in young and older males. The potential neurophysiological mechanisms that underlie that relationship, especially prefrontal oxygenation and vascular health, are presented

Les etudes sur les mouvements de saisie realises en position assise ont montre que la phase du mouvement libre (ou phase de transport) peut etre assimilee a un mouvement de pointage. Ainsi, la duree de cette phase serait en partie fonction de la taille de l'objet a saisir (loi de fitts). Le but de cette etude est d'analyser si ce resultat reste valide quand de nouvelles contraintes sont imposees aux sujets, comme c'est le cas en escalade. Trois experimentations (analyse cinetique, cinematique et de la charge attentionelle) ont ete menees. Les resultats montrent une nette differenciation des effets de posture et des effets de taille de la prise a saisir. Une posture complexe, representant en escalade la contrainte principale, modifie l'organisation cinetique et cinematique du geste de saisie. Les ajustements realises grace a la vision et traditionnellement decrits pour la phase de transport ne sont plus presents quand les grimpeurs experts realisent une saisie. Une preprogrammation plus complete de la phase du mouvement libre permettrait aux exeprts de realiser ce geste sans necessite de feedback. Une fois le contact realise, les informations proprioceptives deviendraient necessaires pour realiser une saisie stable et appliquer les forces fonctionnelles sur la prise. Les grimpeurs experts utiliseraient cette strategie de preprogrammation afin de faciliter le traitement des informations durant la phase instable du geste, et de controler plus efficacement leur situation posturale. Au contraire, les sujets novices utilisent toujours le feedback visuel durant la phase de transport. Durant cette phase, les novices auraient a controler la complexite posturale et les demandes en precision imposees par la tache. Ces resultats mettent en evidence la necessite d'apprehender les situations motrices dans leur complexite. Ils montrent de facon claire que l'organisation et le controle du geste sont modifies selon une organisation hierarchique des contraintes. Enfin, ils confirment que l'evolution du temps de mouvement precisement decrite par fitts (1954) est, au moins en partie, la consequence des ajustements en feedback.

碩士
國立體育大學
運動科學研究所
97
Purpose: The purpose of this study was to compare the differences under different visual condition for dart-throwing. Methods: Twenty-one subjects voluntarily participated this study. In order to obtain throwers’ performance both in segmental motion and dart location, present study used high speed camera (100Hz) to collect data. The study adapted the repeated one-way ANOVA (α＝.05) and qualitative analysis to compare the difference among three different visual condition. Results: The results showed that the dart location of binocular vision were more accuracy and stable than that of advantage vision. The throwing motion of binocular vision and monocular vision weren’t significantly different. The horizontal deviation of dart location of monocular vision was higher than that of binocular vision. The vertical deviation of dart location of monocular vision was also higher than that of binocular vision. Conclusion: The throwing motion of different visual constraint weren’t significantly different, but the information from monocular vision were harmful for judgment including horizontal and vertical direction of judgment. The accuracy and stable of dart location were affected by monocular vision.

碩士
國立臺灣科技大學
工業管理系
106
Light is one of the important factors in a room design. And adequate lighting will give the impression of good scenery and refreshing environment. Watching movie is an activity that is often done in daily life by elder people and youngster. This study examined the best light setting of elder people and youngster during the watching movie task. The total of the subjects are 24 people consisting of 12 elders and 12 youngsters are participated in this study. In this study are examined four different light settings through analyzing ECG and EEG were used to evaluate their physiological signal. Result showed that high color temperature will make people have more visual fatigue during watching movie. The illumination of 500 lx and color temperature of 3000K will make elder people have less visual fatigue. On the other hand, 150 lx of illumination and 3000K of color temperature will make young people have less visual fatigue

博士
國立臺灣科技大學
設計研究所
99
The present research was to investigate the factors that influence final stage of spatial representation development of totally blind person through three studies. The results suggest: (1) the visual condition does not impale the blind personsʹ drawing to develop into a finer and more maturer stage after 18 years old. (2) The participantsʹ drawing performances varied with different tasks and participantsʹ visual conditions. (3) The blind personʹs final stage of the drawing development is similar to the normal personʹs primitive stage; but the order of the development stages is not the same. It is concluded that vision can't be replaced by the tactility from the aspect of perspective. It is consistent with earlier research by I and Shiu and rejects Kennedyʹs theory again.

Thesis (Ph. D.)--Florida State University, 2004.
Advisor: Dr. Gershon Tenenbaum, Florida State University, College of Education, Dept. of Educational Psychology and Learning Systems. Title and description from dissertation home page (viewed June 15, 2004). Includes bibliographical references.

Dissertação de Mestrado em Engenharia Informática apresentada à Faculdade de Ciências e Tecnologia
Atualmente, as manutenções na aviação são maioritariamente pré-programadas ou reactivas. Estas técnicas podem levar a manutenção excessiva, mais tempo e recursos gastos, menor disponibilidade de aeronaves e despesas altas. Real-time Condition-based Maintenance for Adaptive Aircraft Maintenance Planning (ReMAP) pretende solucionar este problema, para isso introduzindo uma proposta que procura, fazendo uso de sensores e algoritmos de última geração, aplicar condition-based maintenance (CBM) em aeronaves comerciais, possibilitando a integração da monitorização em tempo real dos sistemas de aeronaves com um agente reinforcement learning (RL) de tomada de decisões, capaz de formular um plano de manutenções ótimo. A proposta integra várias work packages (WPs), cada uma delas com um papel diferente no projeto: desde a extração de dados dos sistemas das aeronaves à definição de um plano de manutenção para uma frota inteira. Este estágio foi embutido na WP responsável pelo desenvolvimento da MDST, encarregue de planear e atribuir tarefas de manutenção para uma frota de aviões. A versão final desta ferramenta irá receber informação de cada manutenção periódica de cada aeronave juntamente com pedidos de manutenção inesperados que possam surgir dos prognósticos de saúde das aeronaves. O intuito será distribuir essas tarefas por um prazo pré-definido de maneira a criar um calendário de manutenções praticável. Todavia, inicialmente, como prova de conceito, a MDST irá criar um planeamento que apenas inclua manutenções periódicas. Portanto, o objetivo do estágio foi usar RL de maneira a calendarizar trabalhos de manutenção de uma frota com 51 aviões para o prazo de 6 anos. Os nossos resultados demonstram que o algoritmo de RL converge rapidamente e tem um melhor desempenho do que uma solução gananciosa.
Currently, maintenance in aviation is mostly pre-scheduled or done reactively. These techniques may prompt to over-maintenance, more time and resources spent, less aircraft availability and high expenses. Real-time Condition-based Maintenance for Adaptive Aircraft Maintenance Planning (ReMAP) intends to solve this issue by introducing a proposal that seeks, by using sensors and state-of-the-art algorithms, to apply condition-based maintenance (CBM) in commercial aircraft, allowing real-time health monitoring of the aircraft’s systems to be integrated with a decision making reinforcement learning (RL) agent to formulate an optimal maintenance scheduling plan. The proposal integrates different work packages (WPs), each of them with a different role in the project: from collecting data from the aircraft’s systems to defining a maintenance plan for a whole fleet. This internship was embedded in the WP responsible for the development of the maintenance decision support tool (MDST), in charge of planning and assigning maintenance tasks for a fleet of aircraft. The final version of the tool is envisioned to receive information about each aircraft’s periodical maintenance checks plus any unexpected maintenance needs that might arise from the aircraft’s health prognostics. These maintenance tasks will be assigned to time-frames in order to create a feasible maintenance schedule. Nevertheless, as an introductory concept proof, the MDST must organize a maintenance schedule for periodic tasks, only. Accordingly, the internship’s goal was to use RL to assign periodical maintenance checks for a fleet of 51 aircraft over 6 years. Our results show that the RL algorithm converges quickly and performs significantly better than a greedy solution.

碩士
國立成功大學
物理治療學系
104
It is very common to do two tasks at the same time in daily living, such as talking while driving, listening to music while walking. Doing two tasks at the same time is called dual task. Postural control used to be considered as an automatic process. However, many studies have shown that postural control needs attention, especially in the elderly, and persons with Parkinson’s diseases. Such information is limited in children and furthermore less in children with developmental coordination disorder (DCD). DCD refer to the difficulty in motor coordination and such difficulty cannot be explained by any general medical condition. Difficulty of motor skill and high variability of motor performance are main characteristics of children with DCD. Research results of the effects on the standing postural control in children with DCD are inconsistent. Some results show that performing a concurrent cognitive task will impair the postural control in children with DCD, but some show the opposite results. The concurrent cognitive task may be a visual-spatial working memory task, or an auditory working memory task. The effects may also be different. Therefore, the purpose of this research is to examine the effect of a dual task condition on the standing postural control in children with DCD. Twenty children with DCD and 20 typically developing (TD) children participated in the study. Two types of concurrent working memory tasks: visual spatial and auditory, and two kind of standing conditions: standard stance and tandem stance were the experiment tasks. The performance of standing postural control task was expressed with the velocity and trajectory of center of the pressure (COP). The cognitive task performance was expressed by accuracy and reaction time. We used the dual task costs (performance in dual task condition minus performance in single task condition) as the dependent variables for analyses. The results showed that there was group effect of dual task cost, and such effect was affected by standing postures, but not affected by working memory tasks. That is children with DCD showed worse standing stability in tandem stance condition while concurrently performing a working memory task, but TD children had better standing stability while concurrently performing a working memory task. However such effect was not different between performing a visual spatial task and performing an auditory task. In conclusion, children with DCD showed worse standing stability in dual task condition; TD children showed better standing stability in dual task condition. Dual task condition may enhance the attention in TD children, but not in children with DCD.

Pedestrian detection is a core problem in computer vision due to its centrality to a range of applications such as robotics, video surveillance, and advanced driving assistance systems. Despite its broad application and interest, it remains a challenging problem in part due to the vast range of conditions under which it must be robust. In particular, pedestrian detectors must be robust and reliable at nighttime and in adverse weather conditions, which are some reasons why thermal and multispectral approaches have become popular in recent years. Moreover, thermal imagery offers more privacy-preserving affordances than visible-spectrum surveillance images. However, pedestrian detection in the thermal domain remains a non-trivial task with much room for improvement. Thermal detection helps ameliorate some of the disadvantages of RGB detectors -- such as illumination variation and the various complications of detection at nighttime. However, detection using only thermal imagery still faces numerous challenges and overall lack of information in thermal images. Thermal images are typically low-resolution, which in turn leads to more challenging detection of small pedestrians. Finally, there is a general lack of thermal imagery for training state-of-the-art detectors for thermal detection. The best pedestrian detectors available today work in the visible spectrum. In this thesis, we present three new types of domain adaptation approaches for pedestrian detection in thermal imagery and demonstrate how we can mitigate the above challenges such as privacy-preserving, illumination, lacking thermal data for training, and lacking feature information in thermal images and advance the state-of-the-art. Our first contribution is two \emph{bottom-up domain adaptation} approaches. We first show that simple bottom-up domain adaptation strategies with a pre-trained \emph{adapter} segment can better preserve features from source domains when doing transfer learning of pre-trained models to the thermal domain. In a similar vein, we then show that bottom-up and \emph{layer-wise} adaptation consistently results in more effective domain transfer. Experimental results demonstrate efficiency, flexibility, as well as the potential of both bottom-up domain adaptation approaches. Our second contribution, which addresses some limitations of domain adaptation to thermal imagery, is an approach based on task-conditioned networks that simultaneously solve two related tasks. A detection network is augmented with an auxiliary classification pipeline, which is tasked with classifying whether an input image was acquired during the day or at nighttime. The feature representation learned to solve this auxiliary classification task is then used to \emph{condition} convolutional layers in the main detector network. The experimental results of task-conditioned domain adaptation indicate that task conditioning is an effective way to balance the trade-off between the effectiveness of thermal imagery at night and its weaknesses during the day. Finally, our third contribution addresses the acute lack of training data for thermal domain pedestrian detection. We propose an approach using GANs to generate synthetic thermal imagery as a type of generative data augmentation. Our experimental results demonstrate that synthetically generated thermal imagery can be used to significantly reduce the need for massive amounts of annotated thermal pedestrian data. Pedestrian detection in thermal imagery remains challenging. However, in this thesis, we have shown that our bottom-up and layer-wise domain adaptation methods -- especially the proposed task-conditioned network -- can lead to robust pedestrian detection results via using thermal-only representations at detection time. This shows the potential of our proposed methods not only for domain adaptation of pedestrian detectors but also for other tasks. Moreover, our results using generated synthetic thermal images also illustrate the potential of generative data augmentation for domain adaptation to thermal imagery.

Structural health monitoring and building assessment are crucial to acquire structures’ states and maintain their conditions. Besides human-labor surveys that are subjective, time-consuming, and expensive, autonomous image and video analysis is a faster, more efficient, and non-destructive way. This thesis focuses on crack detection from videos, crack segmentation from images, and building assessment from street view images. For crack detection from videos, three approaches are proposed based on local binary pattern (LBP) and support vector machine (SVM), deep convolution neural network (DCNN), and fully-connected network (FCN). A parametric Naïve Bayes data fusion scheme is introduced that registers video frames in a spatiotemporal coordinate system and fuses information based on Bayesian probability to increase detection precision. For crack segmentation from images, the rotation-invariant property of crack is utilized to enhance the segmentation accuracy. The architectures of several approximately rotation-invariant DCNNs are discussed and compared using several crack datasets. For building assessment from street view images, a framework of multiple DCNNs is proposed to detect buildings and predict their attributes that are crucial for flood risk estimation, including founding heights, foundation types (pier, slab, mobile home, or others), building types (commercial, residential, or mobile home), and building stories. A feature fusion scheme is proposed that combines image feature with meta information to improve the predictions, and a task relation encoding network (TREncNet) is introduced that encodes task relations as network connections to enhance multi-task learning.