A robust event-driven approach to always-on object recognition

<div> <div> <div> <p>We propose a neuromimetic architecture able to perform always-on pattern recognition. To achieve this, we extended an existing event-based algorithm [1], which introduced novel spatio-temporal features as a Hierarchy Of Time-Surfaces (HOTS). Built from asynchronous events acquired by a neuromorphic camera, these time surfaces allow to code the local dynamics of a visual scene and to create an efficient event-based pattern recognition architecture. Inspired by neuroscience, we extended this method to increase its performance. Our first contribution was to add a homeostatic gain control on the activity of neurons to improve the learning of spatio-temporal patterns [2]. A second contribution is to draw an analogy between the HOTS algorithm and Spiking Neural Networks (SNN). Following that analogy, our last contribution is to modify the classification layer and remodel the offline pattern categorization method previously used into an online and event-driven one. This classifier uses the spiking output of the network to define novel time surfaces and we then perform online classification with a neuromimetic implementation of a multinomial logistic regression. Not only do these improvements increase consistently the performances of the network, they also make this event-driven pattern recognition algorithm online and bio-realistic. Results were validated on different datasets: DVS barrel [3], Poker-DVS [4] and N-MNIST [5]. We foresee to develop the SNN version of the method and to extend this fully event-driven approach to more naturalistic tasks, notably for always-on, ultra-fast object categorization. </p> </div> </div> </div>

A robust event-driven approach to always-on object recognition

<div> <div> <div> <p>We propose a neuromimetic architecture able to perform always-on pattern recognition. To achieve this, we extended an existing event-based algorithm [1], which introduced novel spatio-temporal features as a Hierarchy Of Time-Surfaces (HOTS). Built from asynchronous events acquired by a neuromorphic camera, these time surfaces allow to code the local dynamics of a visual scene and to create an efficient event-based pattern recognition architecture. Inspired by neuroscience, we extended this method to increase its performance. Our first contribution was to add a homeostatic gain control on the activity of neurons to improve the learning of spatio-temporal patterns [2]. A second contribution is to draw an analogy between the HOTS algorithm and Spiking Neural Networks (SNN). Following that analogy, our last contribution is to modify the classification layer and remodel the offline pattern categorization method previously used into an online and event-driven one. This classifier uses the spiking output of the network to define novel time surfaces and we then perform online classification with a neuromimetic implementation of a multinomial logistic regression. Not only do these improvements increase consistently the performances of the network, they also make this event-driven pattern recognition algorithm online and bio-realistic. Results were validated on different datasets: DVS barrel [3], Poker-DVS [4] and N-MNIST [5]. We foresee to develop the SNN version of the method and to extend this fully event-driven approach to more naturalistic tasks, notably for always-on, ultra-fast object categorization. </p> </div> </div> </div>

Resolving the time course of visual and auditory object categorization

Humans can effortlessly categorize objects, both when they are conveyed through visual images and spoken words. To resolve the neural correlates of object categorization, studies have so far primarily focused on the visual modality. It is therefore still unclear how the brain extracts categorical information from auditory signals. In the current study we used EEG (N=47) and time-resolved multivariate pattern analysis to investigate (1) the time course with which object category information emerges in the auditory modality and (2) how the representational transition from individual object identification to category representation compares between the auditory modality and the visual modality. Our results show that (1) that auditory object category representations can be reliably extracted from EEG signals and (2) a similar representational transition occurs in the visual and auditory modalities, where an initial representation at the individual-object level is followed by a subsequent representation of the objects category membership. Altogether, our results suggest an analogous hierarchy of information processing across sensory channels. However, we did not find evidence for a shared supra-modal code, suggesting that the contents of the different sensory hierarchies are ultimately modality-unique.

The dorsal visual pathway represents object-centered spatial relations for object recognition

Although there is mounting evidence that input from the dorsal visual pathway is crucial for object processes in the ventral pathway, the specific functional contributions of dorsal cortex to these processes remains poorly understood. Here, we hypothesized that dorsal cortex computes the spatial relations among an object's parts — a processes crucial for forming global shape percepts — and transmits this information to the ventral pathway to support object categorization. Using multiple functional localizers, we discovered regions in the intraparietal sulcus (IPS) that were selectively involved in computing object-centered part relations. These regions exhibited task-dependent functional connectivity with ventral cortex, and were distinct from other dorsal regions, such as those representing allocentric relations, 3D shape, and tools. In a subsequent experiment, we found that the multivariate response of posterior IPS, defined on the basis of part-relations, could be used to decode object category at levels comparable to ventral object regions. Moreover, mediation and multivariate connectivity analyses further suggested that IPS may account for representations of part relations in the ventral pathway. Together, our results highlight specific contributions of the dorsal visual pathway to object recognition. We suggest that dorsal cortex is a crucial source of input to the ventral pathway and may support the ability to categorize objects on the basis of global shape.

Color Based Object Categorization Using Histograms of Oriented Hue and Saturation

In the last few years, there has been a lot of interest in making smart components, e.g. robots, able to simulate human capacity of object recognition and categorization. In this paper, we propose a new revolutionary approach for object categorization based on combining the HOG (Histograms of Oriented Gradients) descriptors with our two new descriptors, HOH (Histograms of Oriented Hue) and HOS (Histograms of Oriented Saturation), designed it in the HSL (Hue, Saturation and Luminance) color space and inspired by this famous HOG descriptor. By using the chrominance components, we have succeeded in making the proposed descriptor invariant to all lighting conditions changes. Moreover, the use of this oriented gradient makes our descriptor invariant to geometric condition changes including geometric and photometric transformation. Finally, the combination of color and gradient information increase the recognition rate of this descriptor and give it an exceptional performance compared to existing methods in the recognition of colored handmade objects with uniform background (98.92% for Columbia Object Image Library and 99.16% for the Amsterdam Library of Object Images). For the classification task, we propose the use of two strong and very used classifiers, SVM (Support Vector Machine) and KNN (k-nearest neighbors) classifiers.

Object Categorization of the English Nominations for Animals Diseases

Introduction. The article is devoted to the study of features of object categorization in veterinary terminology for animal diseases. The relevance of the article is due to the need to study the object categorization in the formation, structuring and functioning of the English terminology of veterinary medicine, in particular the terms for animal diseases. The analysis was performed using definitive, semantic, categorical and conceptual research methods. The purpose of the research is to study object categorization of the English terms for animal diseases. Materials and methods of research. English veterinary terms that name animal diseases, obtained by the method of continuous sampling from specialized dictionaries were chosen as the material for research Results of the research. The category OBJECT is widely used in veterinary terminology, as it is one of the basic categories for the classification of concepts that are nominated by terms for animal diseases. Terms representing the category of OBJECT are divided into three groups: 1) terms for affected anatomical objects; 2) terms for pathological objects; 3) terms to indicate the diseases of certain animals. The author represents groups of nominations that correspond to the category OBJECT. Thus, the category of OBJECT is widely used in veterinary terminology, as it is one of the basic categories that play an important role in the formation of a scientific concept, as well as in the formation of the name that reflects it. Conclusions. Object categorization is one of the principles of classification of animal diseases, on the basis of which the process of cognition of objects of veterinary medicine is realized, as well as the formation, structuring and functioning of English terminology for animal diseases. We see the prospect of research in identifying a set of language tools in English that are used to verbalize the category OBJECT in the terminology for animal diseases.

Image Classification Between Two Animals

We are going to examine the fine-grained object categorization problem of identifying the breed of animal from a picture. To this end we introduce a replacement annotated dataset of pets covering 37 different breeds of cats and dogs. The visual problem is extremely challenging for the cat and dog, particularly cats, are very deformable and there are often exactly subtle differences between their breeds. We make variety of contributions: we first introduce a model to classify cat and dog breed automatically from a picture. The model adding the shape of the pet animals, captured by a deformable part model detecting the cat and dog face, and appearance, captured by a bag-of-words model that describes the pet fur. Fitting the model involves automatically segmenting the cat and dog within the image. Second, we compare two classification approaches: a hierarchical one, during which a pet animal is first assigned to the cat or dog family then to a breed, and a flat one, during which the breed is obtained directly.