Search

Deep neural networks perform well in object recognition, but do they perceive objects like humans? This study investigates the Gestalt principle of closure in convolutional neural networks. We propose a protocol to identify closure and conduct experiments using simple visual stimuli with progressively removed edge sections. We evaluate well-known networks on their ability to classify incomplete polygons. Our findings reveal a performance degradation as the edge removal percentage increases, indicating that current models heavily rely on complete edge information for accurate classification. The data used in our study is available on Github., Comment: Published at the ESANN 2024 proceedings, European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning. Bruges (Belgium) and online event, 9-11 October 2024

Assessing the aesthetic quality of artistic images presents unique challenges due to the subjective nature of aesthetics and the complex visual characteristics inherent to artworks. Basic data augmentation techniques commonly applied to natural images in computer vision may not be suitable for art images in aesthetic evaluation tasks, as they can change the composition of the art images. In this paper, we explore the impact of local and global data augmentation techniques on artistic image aesthetic assessment (IAA). We introduce BackFlip, a local data augmentation technique designed specifically for artistic IAA. We evaluate the performance of BackFlip across three artistic image datasets and four neural network architectures, comparing it with the commonly used data augmentation techniques. Then, we analyze the effects of components within the BackFlip pipeline through an ablation study. Our findings demonstrate that local augmentations, such as BackFlip, tend to outperform global augmentations on artistic IAA in most cases, probably because they do not perturb the composition of the art images. These results emphasize the importance of considering both local and global augmentations in future computational aesthetics research., Comment: Published at the VISART VII workshop at ECCV 2024. Ombretta Strafforello, Gonzalo Muradas Odriozola, Fatemeh Behrad, Li-Wei Chen, Anne-Sofie Maerten and Derya Soydaner contributed equally to this work

The human brain has an inherent ability to fill in gaps to perceive figures as complete wholes, even when parts are missing or fragmented. This phenomenon is known as Closure in psychology, one of the Gestalt laws of perceptual organization, explaining how the human brain interprets visual stimuli. Given the importance of Closure for human object recognition, we investigate whether neural networks rely on a similar mechanism. Exploring this crucial human visual skill in neural networks has the potential to highlight their comparability to humans. Recent studies have examined the Closure effect in neural networks. However, they typically focus on a limited selection of Convolutional Neural Networks (CNNs) and have not reached a consensus on their capability to perform Closure. To address these gaps, we present a systematic framework for investigating the Closure principle in neural networks. We introduce well-curated datasets designed to test for Closure effects, including both modal and amodal completion. We then conduct experiments on various CNNs employing different measurements. Our comprehensive analysis reveals that VGG16 and DenseNet-121 exhibit the Closure effect, while other CNNs show variable results. We interpret these findings by blending insights from psychology and neural network research, offering a unique perspective that enhances transparency in understanding neural networks. Our code and dataset will be made available on GitHub.

Over the past two decades, researchers in the field of visual aesthetics have studied numerous quantitative (objective) image properties and how they relate to visual aesthetic appreciation. However, results are difficult to compare between research groups. One reason is that researchers use different sets of image properties in their studies. But even if the same properties are used, the image pre-processing techniques may differ and often researchers use their own customized scripts to calculate the image properties. To provide greater accessibility and comparability of research results in visual experimental aesthetics, we developed an open-access and easy-to-use toolbox (called the 'Aesthetics Toolbox'). The Toolbox allows users to calculate a well-defined set of quantitative image properties popular in contemporary research. The properties include lightness and color statistics, Fourier spectral properties, fractality, self-similarity, symmetry, as well as different entropy measures and CNN-based variances. Compatible with most devices, the Toolbox provides an intuitive click-and-drop web interface. In the Toolbox, we integrated the original scripts of four different research groups and translated them into Python 3. To ensure that results were consistent across analyses, we took care that results from the Python versions of the scripts were the same as those from the original scripts. The toolbox, detailed documentation, and a link to the cloud version are available via Github: https://github.com/RBartho/Aesthetics-Toolbox. In summary, we developed a toolbox that helps to standardize and simplify the calculation of quantitative image properties for visual aesthetics research., Comment: 41 pages, 6 figure

The allure of aesthetic appeal in images captivates our senses, yet the underlying intricacies of aesthetic preferences remain elusive. In this study, we pioneer a novel perspective by utilizing several different machine learning (ML) models that focus on aesthetic attributes known to influence preferences. Our models process these attributes as inputs to predict the aesthetic scores of images. Moreover, to delve deeper and obtain interpretable explanations regarding the factors driving aesthetic preferences, we utilize the popular Explainable AI (XAI) technique known as SHapley Additive exPlanations (SHAP). Our methodology compares the performance of various ML models, including Random Forest, XGBoost, Support Vector Regression, and Multilayer Perceptron, in accurately predicting aesthetic scores, and consistently observing results in conjunction with SHAP. We conduct experiments on three image aesthetic benchmarks, namely Aesthetics with Attributes Database (AADB), Explainable Visual Aesthetics (EVA), and Personalized image Aesthetics database with Rich Attributes (PARA), providing insights into the roles of attributes and their interactions. Finally, our study presents ML models for aesthetics research, alongside the introduction of XAI. Our aim is to shed light on the complex nature of aesthetic preferences in images through ML and to provide a deeper understanding of the attributes that influence aesthetic judgements.

As people's aesthetic preferences for images are far from understood, image aesthetic assessment is a challenging artificial intelligence task. The range of factors underlying this task is almost unlimited, but we know that some aesthetic attributes affect those preferences. In this study, we present a multi-task convolutional neural network that takes into account these attributes. The proposed neural network jointly learns the attributes along with the overall aesthetic scores of images. This multi-task learning framework allows for effective generalization through the utilization of shared representations. Our experiments demonstrate that the proposed method outperforms the state-of-the-art approaches in predicting overall aesthetic scores for images in one benchmark of image aesthetics. We achieve near-human performance in terms of overall aesthetic scores when considering the Spearman's rank correlations. Moreover, our model pioneers the application of multi-tasking in another benchmark, serving as a new baseline for future research. Notably, our approach achieves this performance while using fewer parameters compared to existing multi-task neural networks in the literature, and consequently makes our method more efficient in terms of computational complexity.

Levan is a bacterial homopolysaccharide, which consists of beta-2,6 linked beta-D-fructose monomers. Because of its structural properties and its health promoting effects, levan is a promising functional ingredient for the food, cosmetic and pharma industry. The properties of levan have been reported to be linked to its molecular weight. For a better understanding of how its molecular weight determines its polymer conformation in aqueous solution, levan produced by the food grade acetic acid bacterium Gluconobacter albidus TMW 2.1191 was analysed over a broad molecular weight range using dynamic and static light scattering and viscometry. Levan, with low molecular weight, exhibited a compact random coil structure. As the molecular weight increased, the structure transformed into a compact non-drained sphere. The density of the sphere continued to increase with increasing molecular weight. This resulted in a negative exponent in the Mark-Houwink-Sakurada Plot. For the first time, an increase in molecular density with increasing molecular weight, as determined by a negative Mark-Houwink-Sakurada exponent, could be shown for biopolymers. Our results reveal the unique properties of high-molecular weight levan and indicate the need of further systematic studies on the structure-function relationship of levans for their targeted use in food, cosmetic and pharmaceutical applications.

High quality nuclear data is the most fundamental underpinning for all neutron metrology applications. This paper describes the release of version II of the International Reactor Dosimetry and Fusion File (IRDFF-II) that contains a consistent set of nuclear data for fission and fusion neutron metrology applications up to 60~MeV neutron energy. The library is intended to support: a) applications in research reactors; b) safety and regulatory applications in the nuclear power generation in commercial fission reactors; and c) material damage studies in support of the research and development of advanced fusion concepts. The paper describes the contents of the library, documents the thorough verification process used in its preparation, and provides an extensive set of validation data gathered from a wide range of neutron benchmark fields. The new \mbox{IRDFF-II} library includes 119 metrology reactions, four cover material reactions to support self-shielding corrections, five metrology metrics used by the dosimetry community, and cumulative fission products yields for seven fission products in three different neutron energy regions. In support of characterizing the measurement of the residual nuclei from the dosimetry reactions and the fission product decay modes, the present document lists the recommended decay data, particle emission energies and probabilities for 68 activation products. It also includes neutron spectral characterization data for 29 neutron benchmark fields for the validation of the library contents. The IRDFF-II library and comprehensive documentation is available online at https://www-nds.iaea.org/IRDFF/. Evaluated cross sections can be compared with experimental data and other evaluations at https://www-nds.iaea.org/exfor/endf.htm. The new library is expected to become the international reference in neutron metrology for multiple applications., Comment: Submitted to Nuclear Data Sheets, 109 pages, figures revised

Accelerator-Driven Systems (ADS) could be employed to incinerate minor actinides and so partly contribute to answer the problem of nuclear waste management. An ADS consists of the coupling of a subcritical fast reactor to a particle accelerator via a heavy material spallation target. The on-line reactivity monitoring of such an ADS is a serious issue regarding its safety. In order to study the methodology of this monitoring, zero-power experimentswere undertaken at the GUINEVERE facility within the framework of the FP6-IP-EUROTRANS programme. Such experiments have been under completion within the FREYA FP7 project. The GUINEVERE facility is hosted at the SCK-CEN site in Mol (Belgium). It couples the VENUS-F subcritical fast core with the GENEPI-3C accelerator. The latter delivers a beam of deuterons, which are converted into 14-MeV neutrons via fusion reactions on a tritiated target. This paper presents one of the investigated methods for ADS on-line reactivity monitoring which has to be validated in the program of the FREYA project. It describes the results obtained when Pulsed Neutron Source experiments are analysed using the so called Area Method, in order to estimate the reactivity of a few sub-critical configurations of the VENUS-F reactor, around keff= 0.96. First the GUINEVERE facility is described. Then, following general considerations on the Area method, the results of its application to the neutron population time decrease spectra measured after a pulse by several fission chambers spread out over the whole reactor are discussed. Finally the reactivity values extracted are compared to the static reactivity values obtained using the Modified Source Multiplication (MSM) method., Comment: International Workshop on Technology and Components of Accelerator Driven Systems (TCADS) OECD Nuclear Energy Agency, Nantes : France (2013)

The 37Ar(n_th,alpha)34S and 37Ar(n_th,p)37Cl reactions were studied at the high flux reactor of the ILL in Grenoble. For the 37Ar(n_th,alpha_0) and 37Ar(n_th,p) reaction cross sections, values of (1070+/-80)b and (37+/-4)b, respectively, were obtained. Both values are about a factor 2 smaller than results of older measurements. The observed suppression of the 37(n_th,alpha_1) transition could be verified from theoretical considerations. Finally, evidence was found for the two-step 37Ar(n_th,gamma-alpha) process., Comment: 11 pages, 5 figures, accepted for publication in Nuclear Physics A

The 17O(n_th,alpha)14C reaction cross section was determined at the high flux reactor of the ILL in Grenoble relative to the known 14N(n_th,p)14C cross section. The 17O(n_th,alpha)14C measurements were performed with several highly enriched oxygen gas samples and the flux calibration was done with 14N_2 from the air. This resulted in a precise value of (244+/-7)mb for the 17O(n_th,alpha)14C cross section., Comment: 4 pages, 3 figures, to be published in proceedings of Nuclei in the Cosmos V (1998)

The energy dependent 37Ar(n,p)37Cl and 37Ar(n,alpha)34S reaction cross sections were determined for the first time in an experimental effort involving three large facilities: the cyclotron of the UCL (Louvain-la-Neuve, Belgium) where implanted 37Ar samples were produced; the high flux reactor of the ILL (Grenoble, France) where thermal (n,p), (n,alpha_0) and (n,alpha_1) cross sections of (37+/-4)b, (1070+/-80)b and (290+/-50)mb respectively could be determined, and the GELINA neutron time-of-flight facility of the IRMM (Geel, Belgium) where strong resonances were observed in the keV region., Comment: 4 pages, 3 figures, to appear in proceedings of Nuclei in the Cosmos V, 1998

A study analyzed the prior knowledge state of two university populations: students at the Open University (OU) of the Netherlands and students at the University of Limburg (UL). Analysis of the prior knowledge state was based on an extensive analysis of the literature in relation to theories, models, and practice-based strategies about the structure of knowledge. A set of four dimensions--cognitive psychological, educational psychological, psychometrical, and content based--that help to construct knowledge profiles was defined. The dimensions were used to analyze the knowledge profiles of the two populations. A knowledge state test that covered the whole domain of economics in the course "Economics and Money" was administered to 626 students; 91 UL students were randomly selected to be compared with 91 OU students. The overall economics score was not significantly different between the two populations. This score was elaborated upon by grouping items along knowledge profile dimensions. After outlining the theoretical base of these dimensions, researchers focused on the application of these dimensions to compare the knowledge profiles of the two populations. Up to 7 of the 10 knowledge profile dimensions proved to be relevant. The content and epistemological dimensions were especially helpful to describe differences. (Contains 28 references.) (YLB)

This study sought to determine the effects of prior knowledge on the achievement of undergraduates in an economics course at the Open University of the Netherlands (OuN). A total the four law and seven economics students were given four instruments to test prior knowledge of economics, presented with a learning unit from the course "Economics and Money," and then post-tested on subject mastery. The results indicated that the second general measure of expertise, comprising the requisite test scores and the test scores for the subject-oriented expertise test, explained up to 42 percent of the variance in the posttest scores. Further analysis revealed that optimal requisite knowledge was an important component of student expertise. The study also found that the estimation of expertise level through self-assessment by the students was not very reliable. It is suggested that other dimensions, such as psychological and epistemological dimensions, can be put forward to analyze expertise. (Contains 20 references.) (MDM)

Analyzes prior knowledge based on the idea of knowledge structure; the relevance of using knowledge profiles is important in independent, distance learning. Suggests that prior knowledge can be helpful to enhance a student's learning process, to design more flexible learning environments, and to provide institutional support. (Author/AEF)