Large Set Of Objects Research Articles

Learning from an Infant’s Visual Experience

Infants see a selective view of the world: they see some objects with high frequency and from a wide range of viewpoints (e.g., their toys during playing) while a much larger set of objects are seen much more rarely and from limited viewpoints (e.g., objects they see outdoors). Extensive, repeated visual experiences with a small number of objects during infancy plays a big role in the development of human visual skills. Internet-style datasets that are commonly used in computer vision research do not contain the regularities that result from such repeated, structured experiences with a few objects. This has led to a dearth of models that learn by exploiting these regularities. In my PhD dissertation, I use deep learning models to investigate how regularities in an infant's visual experience can be leveraged for visual representation learning.

Human shape representations are not an emergent property of learning to classify objects.

Humans are particularly sensitive to relationships between parts of objects. It remains unclear why this is. One hypothesis is that relational features are highly diagnostic of object categories and emerge as a result of learning to classify objects. We tested this by analyzing the internal representations of supervised convolutional neural networks (CNNs) trained to classify large sets of objects. We found that CNNs do not show the same sensitivity to relational changes as previously observed for human participants. Furthermore, when we precisely controlled the deformations to objects, human behavior was best predicted by the number of relational changes while CNNs were equally sensitive to all changes. Even changing the statistics of the learning environment by making relations uniquely diagnostic did not make networks more sensitive to relations in general. Our results show that learning to classify objects is not sufficient for the emergence of human shape representations. Instead, these results suggest that humans are selectively sensitive to relational changes because they build representations of distal objects from their retinal images and interpret relational changes as changes to these distal objects. This inferential process makes human shape representations qualitatively different from those of artificial neural networks optimized to perform image classification. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Scene context is predictive of unconstrained object similarity judgments

What makes objects alike in the human mind? Computational approaches for characterizing object similarity have largely focused on the visual forms of objects or their linguistic associations. However, intuitive notions of object similarity may depend heavily on contextual reasoning—that is, objects may be grouped together in the mind if they occur in the context of similar scenes or events. Using large-scale analyses of natural scene statistics and human behavior, we found that a computational model of the associations between objects and their scene contexts is strongly predictive of how humans spontaneously group objects by similarity. Specifically, we learned contextual prototypes for a diverse set of object categories by taking the average response of a convolutional neural network (CNN) to the scene contexts in which the objects typically occurred. In behavioral experiments, we found that contextual prototypes were strongly predictive of human similarity judgments for a large set of objects and rivaled the performance of models based on CNN representations of the objects themselves or word embeddings for their names. Together, our findings reveal the remarkable degree to which the natural statistics of context predict commonsense notions of object similarity.

Simple kinesthetic haptics for object recognition

Object recognition is an essential capability when performing various tasks. Humans naturally use either or both visual and tactile perception to extract object class and properties. Typical approaches for robots, however, require complex visual systems or multiple high-density tactile sensors which can be highly expensive. In addition, they usually require actual collection of a large dataset from real objects through direct interaction. In this paper, we propose a kinesthetic-based object recognition method that can be performed with any multi-fingered robotic hand in which the kinematics is known. The method does not require tactile sensors and is based on observing grasps of the objects. We utilize a unique and frame invariant parameterization of grasps to learn instances of object shapes. To train a classifier, training data is generated rapidly and solely in a computational process without interaction with real objects. We then propose and compare between two iterative algorithms that can integrate any trained classifier. The classifiers and algorithms are independent of any particular robot hand and, therefore, can be exerted on various ones. We show in experiments, that with few grasps, the algorithms acquire accurate classification. Furthermore, we show that the object recognition approach is scalable to objects of various sizes. Similarly, a global classifier is trained to identify general geometries (e.g., an ellipsoid or a box) rather than particular ones and demonstrated on a large set of objects. Full scale experiments and analysis are provided to show the performance of the method.

Automated Design of Embedded Constraints for Soft Hands Enabling New Grasp Strategies

Soft robotic hands allow to fully exploit hand-object-environment interactions to complete grasping tasks. However, their usability can still be limited in some scenarios (e.g., restricted or cluttered spaces). In this article, we propose to enhance the versatility of soft grippers by adding special passive components to their structure, without completely altering their design, nor their control. A method for the automated design of soft-rigid scoop-shaped add-ons acting as “embedded constraints” is presented. Given a certain gripper and a large set of objects, the design parameters of the optimal scoop for each object are derived by solving an optimization problem. Also the object-environment relative pose is considered in the optimization. The obtained “optimal scoops” are clustered to get a limited set of representative scoop designs which can be prototyped and used in grasping tasks. In this article, we also introduce a data-driven method allowing a grasp planner to select the most suitable scoop to be added to the used hand, given a certain object and its configuration with respect to the surrounding environment. Experiments with two different hands validate the proposed approach.

Bayesian Modeling of Sequential Discoveries

We aim at modeling the appearance of distinct tags in a sequence of labeled objects. Common examples of this type of data include words in a corpus or distinct species in a sample. These sequential discoveries are often summarized via accumulation curves, which count the number of distinct entities observed in an increasingly large set of objects. We propose a novel Bayesian method for species sampling modeling by directly specifying the probability of a new discovery, therefore, allowing for flexible specifications. The asymptotic behavior and finite sample properties of such an approach are extensively studied. Interestingly, our enlarged class of sequential processes includes highly tractable special cases. We present a subclass of models characterized by appealing theoretical and computational properties, including one that shares the same discovery probability with the Dirichlet process. Moreover, due to strong connections with logistic regression models, the latter subclass can naturally account for covariates. We finally test our proposal on both synthetic and real data, with special emphasis on a large fungal biodiversity study in Finland. Supplementary materials for this article are available online.

Reconfigurable Intelligent Surface Enabled Vehicular Communication: Joint User Scheduling and Passive Beamforming

Given its ability to control and manipulate wireless environments, reconfigurable intelligent surface (RIS), also known as intelligent reflecting surface (IRS), has emerged as a key enabler technology for the six-generation (6G) cellular networks. In the meantime, vehicular environment radio propagation is negatively influenced by a large set of objects that cause transmission distortion such as high buildings. Therefore, this work is devoted to explore the area of RIS technology integration with vehicular communications while considering the dynamic nature of such communication environment. Specifically, we provide a system model where RoadSide Unit (RSU) leverages RIS to provide indirect wireless transmissions to disconnected areas, known as dark zones. Dark zones are spots within RSU coverage where the communication links are blocked due to the existence of blockages. In details, a discrete RIS is utilized to provide communication links between the RSU and the vehicles passing through out-of-service zones. Therefore, the joint problem of RSU resource scheduling and RIS passive beamforming or phase-shift matrix is formulated as an optimization problem with the objective of maximizing the minimum average bit rate. The formulated problem is mixed integer non-convex program which is difficult to be solved and does not account for the uncertain dynamic environment in vehicular networks. Thereby, we resort to alternative methods based on Deep Reinforcement Learning to determine RSU wireless scheduling and Block Coordinate Descent (BCD) to solve for the phase-shift matrix, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e.,</i> passive beamforming, of the RIS. The Markov Decision Process (MDP) is defined and the complexity of the solution approach is discussed. Our numerical results demonstrate the superiority of our proposed approach over baseline techniques.

Thermophysical Investigation of Asteroid Surfaces. I. Characterization of Thermal Inertia

Abstract The thermal inertia of an asteroid is an indicator of the thermophysical properties of the regolith and is determined by the size of grains on the surface. Previous thermophysical modeling studies of asteroids have identified or suggested that object size, rotation period, and heliocentric distance (a proxy for temperature) are important factors that separately influence thermal inertia. In this work we present new thermal inertia values for 239 asteroids and model all three factors in a multivariate model of thermal inertia. Using multiepoch infrared data of this large set of objects observed by WISE, we derive the size, albedo, thermal inertia, surface roughness, and sense of spin using a thermophysical modeling approach that does not require a priori knowledge of an object’s shape or spin axis direction. Our thermal inertia results are consistent with previous values from the literature for similarly sized asteroids, and we identify an excess of retrograde rotators among main-belt asteroids &lt;8 km. We then combine our results with thermal inertias of 220 objects from the literature to construct a multivariate model and quantify the dependency on asteroid diameter, rotation period, and surface temperature. This multivariate model, which accounts for codependencies between the three independent variables, identifies asteroid diameter and surface temperature as strong controls on thermal inertia.

Additive trees for the categorization of a large number of objects, with bootstrapping strategy for stability assessment. Application to the free sorting of wine odor terms

In the field of clustering techniques, little attention has been paid to the recovery of a set of clusters from the structure of an additive tree. To bridge this gap, this work presents an original partitioning technique which aims to reveal clusters from an additive tree that represents a large set of objects. Specifically, an algorithm that splits a tree into successive subtrees was developed, based on a ratio of the lengths of edges. The stability of the clusters obtained with this technique was then evaluated using measurements of cohesion and isolation that were generated using a bootstrapping strategy. Finally, the degree of association of each object to clusters was analyzed to gain insight into their internal structure. This analysis was performed on the results of a sorting task conducted by 156 subjects, who were asked to sort 96 terms associated with the odor of wine. The methodology developed in this paper represents an innovative way to highlight groups of terms within a large set of wine odor attributes, with the ultimate goal being to improve the structure of the lexicon.

A map of object space in primate inferotemporal cortex.

The inferotemporal (IT) cortex is responsible for object recognition, but it is unclear how the representation of visual objects is organized in this part of the brain. Areas that are selective for categories such as faces, bodies, and scenes have been found1-5, but large parts of IT cortex lack any known specialization, raising the question of what general principle governs IT organization. Here we used functional MRI, microstimulation, electrophysiology, and deep networks to investigate the organization of macaque IT cortex. We built a low-dimensional object space to describe general objects using a feedforward deep neural network trained on object classification6. Responses of IT cells to a large set of objects revealed that single IT cells project incoming objects onto specific axes of this space. Anatomically, cells were clustered into four networks according to the first two components of their preferred axes, forming a map of object space. This map was repeated across three hierarchical stages of increasing view invariance, and cells that comprised these maps collectively harboured sufficient coding capacity to approximately reconstruct objects. These results provide a unified picture of IT organization in which category-selective regions are part of a coarse map of object space whose dimensions can be extracted from a deep network.

МЕТОДОЛОГІЧНІ ЗАСАДИ КЛАСИФІКАЦІЇ ІНСТРУМЕНТІВ ЦИФРОВИХ СИСТЕМ «SMART CITY»

In the last decade, the development and integration of the life of information technology society has affected the system of public governance. Key aspects of organizing modern urban development, management and planning that encompass different measurements: effective management, stability, environmental friendliness, innovation and understanding of the «Smart City» concept. «Smart City» is a digital toolkit (system), in which the city аdministration interact with the community and the city's infrastructure. As the basic tools of «Smart City» we can consider software and hardware complexes that are the part of the components (applications, software solutions, dedicated services, etc.), that perform specific typical operations for solving certain individual functions and form the overall architecture of the system. A significant barrier for the development of digital interference in Ukraine is a lack of clear classification of «Smart City» warehouse tools. One of the most important ways for systematization is the general classification of «Smart City» tools. The purpose of the article is the theoretical and methodological substantiation of the classification of «Smart City» tools. The article proposes to define the classification method as a set of rules for the creation of classification systems, as well as their coincidence. The proposed facet method most closely matches the system of classification of «Smart City» tools and allows to systematize a large set of objects. It is substantiated that using the facet method it is possible to construct a classification system of «Smart City» tools with flexible scaling at each level of such structure. The structure of classification of basic functions of public administration bodies is determined. The basic digital interactions of public intervention agencies (G - government) with the company (B - business) and citizens (C - citizen) are selected based on the levels of interaction of subjects and objects of electronic service provision. This will make it possible to formulate the first basic classification level depending on a certain initial position of objects in system. In order to determine the interaction levels, it is suggested to introduce index of the «Smart City» solutions, with the positioning of the system component by the classifiers. Based on the facet principle and defined types of fields, the basic system of classification «Smart City» tools was built. Having defined the tool of classification it becomes possible to describe and formaloze the inter-system interconnections at the level of data exchange between different components and subsystems. It expands the possibilities of implementation, cloning and scaling up successful technical, technological and managerial decisions. The proposed classification is a supersystem and does not affect internal architecture of specific systems and is appointmed, first of all, for the construction and standardization of inter-system interactions and systematization of «Smart City» tools. As a further development of issues we see the development of standardization and «Smart City» tools at the state level, the study of the world's experience in the implementation of «Smart City» to determine the recommendations for the development and construction of such systems.

Reduction of the size of L-fuzzy contexts. A tool for differential diagnoses of diseases

ABSTRACTInformation extraction from an L-fuzzy context becomes a hard problem when we work with a large set of objects and/or attributes. The goal of this paper is to present two different and complementary techniques to reduce the size of the context. First, using overlap indexes, we will establish rankings among the elements of the context that will allow us to determine those that do not provide relevant information and eliminate them. Second, by means of Choquet integrals, we will aggregate some objects or attributes of the context in order to jointly use the provided information. One interesting application of the developed theory consists on helping in the differential diagnoses of diseases that share a large number of symptoms and, therefore, that are difficult of distinguish.

Dynamic “Molecular Portraits” of Proteins and Cell Membranes: A Computational View

Molecular surfaces are the key players in processes of bimolecular recognition, interaction, and signaling. Nowadays, state of the art methods exist for visualizing molecule surface and surface distributed properties in three-dimensional space. However, such a visual information could only be analyzed by human eye and therefore prompt to be biased and onerous in case of biomacromolecules with a complex structural organization or for large sets of objects. This is especially true for such mesoscopic systems like cell membranes. Therefore, alternative (mostly computational) techniques operating with such “molecular portraits” are indispensable. Here, we present a number of original in silico methods to map and process multivariate and dynamic surface properties of biomolecules and their assemblies - proteins, membranes, carbohydrates and their complexes. These techniques were shown to improve protein-ligand [1] and protein-protein [2] docking results. A new protein surface topography (PST) method [3] is described. It permits pictorial visualization of the whole protein surface in terms of 2D “earth maps”. PST can be used to delineate conformational changes, perform group analysis of molecules. It is a useful tool to add to docking experiments. The proposed original tools have been shown to be very efficient in analysis of atomistic details of the recognition processes in cell membranes - they allow easy monitoring of dynamic amphiphilic properties of lipid membranes, proteins, peptides, polysaccharides, etc. Implementation of the “molecular portrait” technologies is capable of rational design of new physiologically important compounds with predefined activities [4]. References. 1. Pyrkov T.V. et al. (2009) Bioinformatics. 25:1201. 2. Polyansky A.A. et al. (2014) Bioinformatics 30:889. 3. Koromyslova A.D. et al. (2014) J. Chem. Inf. Mod. 54:1189. 4. Kasheverov I.E. et al. (2016) Scientific Reports, 6:36848. Acknowledgements Russian Foundation for Basic Research (18-54-15007), Russian Science Foundation (18-14-00375).

Robust Tactile Descriptors for Discriminating Objects From Textural Properties via Artificial Robotic Skin

In this paper, we propose a set of novel tactile descriptors to enable robotic systems to extract robust tactile information during tactile object explorations, regardless of the number of the tactile sensors, sensing technologies, type of exploratory movements, and duration of the objects’ surface exploration. The performance and robustness of the tactile descriptors are verified by testing on four different sensing technologies (dynamic pressure sensors, accelerometers, capacitive sensors, and impedance electrode arrays) with two robotic platforms (one anthropomorphic hand and one humanoid), and with a large set of objects and materials. Using our proposed tactile descriptors, the Shadow Hand, which has multimodal robotic skin on its fingertips, successfully classified 120 materials (100% accuracy) and 30 in-hand objects (98% accuracy) with regular and irregular textural structure by executing human-like active exploratory movements on their surface. The robustness of the proposed descriptors was assessed further during the large object discrimination with a humanoid. With a large sensing area on its upper body, the humanoid classified 120 large objects with multiple weights and various textures while the objects slid between its sensitive hands, arms, and chest. The achieved 90% recognition rate shows that the proposed tactile descriptors provided robust tactile information from the large number of tactile signals for identifying large objects via their surface texture regardless of their weight.

A Survey on Data‐driven Dictionary‐based Methods for 3D Modeling

AbstractDictionaries are very useful objects for data analysis, as they enable a compact representation of large sets of objects through the combination of atoms. Dictionary‐based techniques have also particularly benefited from the recent advances in machine learning, which has allowed for data‐driven algorithms to take advantage of the redundancy in the input dataset and discover relations between objects without human supervision or hard‐coded rules. Despite the success of dictionary‐based techniques on a wide range of tasks in geometric modeling and geometry processing, the literature is missing a principled state‐of‐the‐art of the current knowledge in this field. To fill this gap, we provide in this survey an overview of data‐driven dictionary‐based methods in geometric modeling. We structure our discussion by application domain: surface reconstruction, compression, and synthesis. Contrary to previous surveys, we place special emphasis on dictionary‐based methods suitable for 3D data synthesis, with applications in geometric modeling and design. Our ultimate goal is to enlight the fact that these techniques can be used to combine the data‐driven paradigm with design intent to synthesize new plausible objects with minimal human intervention. This is the main motivation to restrict the scope of the present survey to techniques handling point clouds and meshes, making use of dictionaries whose definition depends on the input data, and enabling shape reconstruction or synthesis through the combination of atoms.

Guiding Exploratory Behaviors for Multi-Modal Grounding of Linguistic Descriptions

A major goal of grounded language learning research is to enable robots to connect language predicates to a robot's physical interactive perception of the world. Coupling object exploratory behaviors such as grasping, lifting, and looking with multiple sensory modalities (e.g., audio, haptics, and vision) enables a robot to ground non-visual words like ``heavy'' as well as visual words like ``red''. A major limitation of existing approaches to multi-modal language grounding is that a robot has to exhaustively explore training objects with a variety of actions when learning a new such language predicate. This paper proposes a method for guiding a robot's behavioral exploration policy when learning a novel predicate based on known grounded predicates and the novel predicate's linguistic relationship to them. We demonstrate our approach on two datasets in which a robot explored large sets of objects and was tasked with learning to recognize whether novel words applied to those objects.

Enumeration Processes under Attack: The Role of Working Memory in Subitizing and Serial Counting

Enumeration of visual stimuli is supported by two distinct psychological processes: serial counting and subitizing. Serial counting is an effortful, slow, and controlled process employed for the enumeration of large sets of objects. Subitizing is defined as a fast and accurate assessment of small quantities. Over the past two decades, there has been an on-going debate on whether subitizing and serial counting are based on shared or distinct cognitive mechanisms. Recent theories suggest that subitizing is supported by visual skills related to perception while serial counting requires working memory. The current study examines the respective roles of phonological and spatial working memory in enumeration processes. The main task used was an enumeration task, in which participants named the quantities of randomly arranged dots in the subitizing (1-3 dots) and counting (7-9 dots) ranges. The performance in the enumeration naming task was compared to a dual-task setting in which participants performed the enumeration naming task while they retained a phonological load or a spatial load. Load type had differential impact on enumeration processes. Importantly, it was found that phonological load, but not spatial load, decreased the effectiveness of serial counting. Subitizing ability was not affected by spatial or phonological load. In line with the majority of previous studies on enumeration, our results indicate that enumeration of small and large quantities are based on different cognitive processes. Moreover, the present finding demonstrate that, phonological working memory plays a significant role in serial counting but not in subitizing and that spatial load is not involved in enumeration.

Feature fragments and evidence accumulation in object and face perception

Recognition and identification of objects is ubiquitous in the daily lives of human beings, yet despite its importance we still do not understand how visual information arriving at the retina is processed by the human visual cortex (and the rest of the brain) in such a fast and flexible manner. There is a large body of neurophysiological work with non-human primates showing that certain neurons in the posterior and middle regions of the inferior temporal gyrus (area TE) respond preferentially to specific features of objects. Further studies exploring the brain’s response to a large set of objects decomposed into simpler shapes led a number of researchers to suggest that part of TE was organized in column-like structures with all neurons in a column having preferential tuning for specific shapes. Assuming that object features are presented similarly in humans, the current study aimed to discover the features of objects and faces that produce the fastest and most accurate categorization responses. Participants viewed many single fragments selected from random locations in images of a set of objects and attempted to classify the fragment based on object category. Accuracy and reaction time measures were analyzed using a sequential sampling framework, specifically Drift Diffusion and Linear Ballistic Accumulator models (Brown & Heathcote, 2008; Ratcliff & Rouder, 1998). Modeling results showed that the variability in reaction times across features was accounted by the rate of perceptual evidence accumulation, which is dependent on the amount of information in the fragment. The current study represents a fundamental step toward evaluating the functional organization of the human ventral visual stream within a feature-based theoretical framework. Meeting abstract presented at VSS 2015

High-Performance Annotation Tagging over Solr Full-text Indexes

&lt;p class="p1"&gt;In this work, we focus on the problem of “annotation tagging” over Information &lt;span style="font-size: 10px;"&gt;Spaces of objects stored in a full-text index. In such a scenario, tags are &lt;/span&gt;&lt;span style="font-size: 10px;"&gt;assigned to objects by “data curator” users with the purpose of classification, while &lt;/span&gt;&lt;span style="font-size: 10px;"&gt;generic end-users will perceive tags as searchable and browsable object properties. &lt;/span&gt;&lt;span style="font-size: 10px;"&gt;To carry out their activities, data curators need “annotation tagging tools” which &lt;/span&gt;&lt;span style="font-size: 10px;"&gt;allow them to “bulk” tag or untag large sets of objects in temporary work sessions, &lt;/span&gt;&lt;span style="font-size: 10px;"&gt;where they can “virtually” and in “real-time” experiment the effect of their actions &lt;/span&gt;&lt;span style="font-size: 10px;"&gt;before making the changes visible to end-users. The implementation of these tools &lt;/span&gt;&lt;span style="font-size: 10px;"&gt;over full-text indexes is a challenge, since bulk object updates in this context are &lt;/span&gt;&lt;span style="font-size: 10px;"&gt;far from being real-time and in critical cases may slow down index performance. &lt;/span&gt;&lt;span style="font-size: 10px;"&gt;We devised TagTick, a tool which offers to data curators a fully functional annotation &lt;/span&gt;&lt;span style="font-size: 10px;"&gt;tagging environment over the full-text index Apache Solr, regarded as a &lt;/span&gt;&lt;span style="font-size: 10px;"&gt;“de-facto standard” in this area. TagTick consists of a TagTick Virtualizer module, &lt;/span&gt;&lt;span style="font-size: 10px;"&gt;which extends the APIs of Solr to support real-time, virtual, bulk-tagging operations, &lt;/span&gt;&lt;span style="font-size: 10px;"&gt;and a TagTick User Interface module, which offers end-user functionalities &lt;/span&gt;&lt;span style="font-size: 10px;"&gt;for annotation tagging. The tool scales optimally with the number and size of bulk &lt;/span&gt;&lt;span style="font-size: 10px;"&gt;tag operations, without compromising index performance.&lt;/span&gt;&lt;/p&gt;

Spectroscopic needs for imaging dark energy experiments

Large sets of objects with spectroscopic redshift measurements will be needed for imaging dark energy experiments to achieve their full potential, serving two goals:_training_, i.e., the use of objects with known redshift to develop and optimize photometric redshift algorithms; and_calibration_, i.e., the characterization of moments of redshift (or photo-z error) distributions. Better training makes cosmological constraints from a given experiment stronger, while highly-accurate calibration is needed for photo-z systematics not to dominate errors. In this white paper, we investigate the required scope of spectroscopic datasets which can serve both these purposes for ongoing and next-generation dark energy experiments, as well as the time required to obtain such data with instruments available in the next decade. Large time allocations on kilo-object spectrographs will be necessary, ideally augmented by infrared spectroscopy from space. Alternatively, precision calibrations could be obtained by measuring cross-correlation statistics using samples of bright objects from a large baryon acoustic oscillation experiment such as DESI. We also summarize the additional work on photometric redshift methods needed to prepare for ongoing and future dark energy experiments.