A cost-efficient face learning mechanism: The impact of stability in appearance on the resolution of facial representations

The way faces become familiar and what information is represented as familiarity develops has puzzled researchers in the field of human face recognition for decades. In this paper, we propose a cost-efficient mechanism of face learning to describe how facial representations form over time and that explains why recognition errors occur. Encoding of diagnostic facial information would follow a coarse-to-fine trajectory, modulated by the intrinsic stability in individual faces’ appearance. In four experiments, we draw on a robust and ecological method using a proxy of exposure to famous faces in the real world to test hypotheses generated by the model and we manipulate test images to probe the nature of facial representations. We consistently show that stable facial appearances help create more reliable representation in early stages of familiarisation but that their resolution remains relatively low and therefore less discriminative over time. In contrast, variations in appearance hinder recognition at first but encourage refinement of representations with further exposure. Consistent with the cost-efficient face learning mechanism we propose, facial representations built on a foundation of large-scale coarse information. When coarse information loses its diagnostic value through the experience of variations across encounters, facial details and their spatial relationships receive additional representational weights.

Diverse routes to expertise in facial recognition

Facial recognition errors jeopardize national security, criminal justice, public safety and civil rights. Here, we compare the most accurate humans and facial recognition technology in a detailed lab-based evaluation and international proficiency test for forensic scientists involving 27 forensic departments from 14 countries. We find striking cognitive and perceptual diversity between naturally skilled super-recognizers, trained forensic examiners and deep neural networks, despite them achieving equivalent accuracy. Clear differences emerged in super-recognizers’ and forensic examiners’ perceptual processing, errors, and response patterns: super-recognizers were fast, biased to respond ‘same person’ and misidentified people with extreme confidence, whereas forensic examiners were slow, unbiased and strategically avoided misidentification errors. Further, these human experts and algorithms disagreed on the similarity of faces, pointing to differences in their face representations. Our findings reveal there are multiple types of facial recognition expertise, some of which are better suited to particular real-world facial recognition roles than others.

Identification Performance During Quarantine by COVID-19 Pandemic: Influence of Emotional Variables and Sleep Quality

The COVID-19 pandemic has caused major disruptions in people’s lives around the globe. Sleep habits and emotional balance have been disturbed in a way that could be comparable to the havoc caused by a deep personal crisis or a traumatic experience. This unfortunate situation provides a unique context in which to study the impact of these imbalances on cognitive processes. In particular, the field of eyewitness science could benefit from these conditions, since they are also often present in crime victims, but can only be generated in the laboratory up to a certain ethical and practical limit. For several decades, eyewitness studies have tried to discover what variables affect people’s ability to properly recognize faces. However, the disparity of experimental designs and the limitations of laboratory work could be contributing to the lack of consensus around several factors, such as sleep, anxiety, and depression. Therefore, the possibility of observing the influence of these agents in natural contexts could shed light on this discussion. Here, we perform simple and repeated lineups with witnesses of mock-crime, considering the conditions related to the COVID-19 pandemic, which to some extent allow emulating the deterioration in general well-being that often afflicts crime victims. For this, 72 participants completed symptomatology scales, and watched a video portraying a staged violent episode. Subsequently, they gave testimony and participated in two lineups, in which we manipulated the presence/absence of the perpetrator, to recreate critical scenarios for the appearance of false recognitions. We found an increase in recognition errors in those individuals who did not have access to the perpetrator during the Initial lineup. Additionally, the conditions of the pandemic appear to have adversely affected the ability to witness and accurately perform lineups. These results reaffirm the need to move toward the standardization of research practices and methods for assessing testimonial evidence, especially in relation to the results of the lineups. Considering the degree of fallibility of these processes can lead to a reduction of wrongful convictions.

Multi-scale CNN Approach for Accurate Detection of Underwater Static Fish Image

Recently, the identification and naming of fish species in underwater imagery processing has been in high demand. This is an essential activity for everyone, from biologists to scientists to fisherman. Humans' interests have recently expanded from the earth to the sky and the sea. Robots could be utilized to send mankind to explore the ocean and outer space, as well as for some dangerous professions that human beings are unlikely to perform. Humans have recently shifted their focus from land-based exploration to celestial exploration and the sea. Robots are used for the activities that pose a risk to mankind, like exploration of the seas and outer space. This research article provides a solution to underwater image detection techniques by using an appended transmission map, refinement method and deep learning approach. The features are deeply extracted by multi-scale CNN for attaining higher accuracy in detecting fish features from the input images with the help of segmentation process. Object recognition errors are minimized and it has been compared with other traditional processes. The overall performance metrics graph has been plotted for the proposed algorithm in the results and discussion section.

GPS-Free, Error Tolerant Path Planning for Swarms of Micro Aerial Vehicles with Quality Amplification

We present an error tolerant path planning algorithm for Micro Aerial Vehicle (MAV) swarms. We assume navigation without GPS-like techniques. The MAVs find their path using sensors and cameras, identifying and following a series of visual landmarks. The visual landmarks lead the MAVs towards their destination. MAVs are assumed to be unaware of the terrain and locations of the landmarks. They hold a priori information about landmarks, whose interpretation is prone to errors. Errors are of two types, recognition or advice. Recognition errors follow from misinterpretation of sensed data or a priori information, or confusion of objects, e.g., due to faulty sensors. Advice errors are consequences of outdated or wrong information about landmarks, e.g., due to weather conditions. Our path planning algorithm is cooperative. MAVs communicate and exchange information wirelessly, to minimize the number of recognition and advice errors. Hence, the quality of the navigation decision process is amplified. Our solution successfully achieves an adaptive error tolerant navigation system. Quality amplification is parameterized with respect to the number of MAVs. We validate our approach with theoretical proofs and numeric simulations.

Multiteam Competitive Optimization Algorithm and Its Application in Bearing Fault Diagnosis

A novel multiteam competitive optimization (MTCO) algorithm has been proposed to diagnose the fault patterns of bearings. This algorithm is inspired by competitive behaviors of multiple teams. It is a three-level organization structure; thus, more potential optimal areas can be searched. By imitating human thinking, such as the betrayal and replacement behavior along with the introduction of an acceptable vector, new strategies within the MTCO are designed to increase the diversity and guide jumping out of location suboptimal areas. In addition to this, a kernel function has been introduced to reduce the recognition errors caused by data which are nonlinearly distributed in original space. The obtained experimental results demonstrate that the proposed MTCO is globally stable and optimal decision performance. After that the MTCO is applied for the fault diagnosis of bearings, and it has also been compared with other commonly used methods. The comparison indicates that the proposed algorithm has higher recognition accuracy.

The problem of loss of solutions in the task of searching similar documents: Applying terminology in the construction of a corpus vector model

This article considers the problem of finding text documents similar in meaning in the corpus. We investigate a problem arising when developing applied intelligent information systems that is non-detection of a part of solutions by the TF-IDF algorithm: one can lose some document pairs that are similar according to human assessment, but receive a low similarity assessment from the program. A modification of the algorithm, with the replacement of the complete vocabulary with a vocabulary of specific terms is proposed. The addition of thesauri when building a corpus vector model based on a ranking function has not been previously investigated; the use of thesauri has so far been studied only to improve topic models. The purpose of this work is to improve the quality of the solution by minimizing the loss of its significant part and not adding “false similar” pairs of documents. The improvement is provided by the use of a vocabulary of specific terms extracted from the text of the analyzed documents when calculating the TF-IDF values for corpus vector representation. The experiment was carried out on two corpora of structured normative and technical documents united by a subject: state standards related to information technology and to the field of railways. The glossary of specific terms was compiled by automatic analysis of the text of the documents under consideration, and rule-based NER methods were used. It was demonstrated that the calculation of TF-IDF based on the terminology vocabulary gives more relevant results for the problem under study, which confirmed the hypothesis put forward. The proposed method is less dependent on the shortcomings of the text layer (such as recognition errors) than the calculation of the documents’ proximity using the complete vocabulary of the corpus. We determined the factors that can affect the quality of the decision: the way of compiling a terminology vocabulary, the choice of the range of n-grams for the vocabulary, the correctness of the wording of specific terms and the validity of their inclusion in the glossary of the document. The findings can be used to solve applied problems related to the search for documents that are close in meaning, such as semantic search, taking into account the subject area, corporate search in multi-user mode, detection of hidden plagiarism, identification of contradictions in a collection of documents, determination of novelty in documents when building a knowledge base.

Advanced atomic force microscopy-based techniques for nanoscale characterization of switching devices for emerging neuromorphic applications

Neuromorphic systems require integrated structures with high-density memory and selector devices to avoid interference and recognition errors between neighboring memory cells. To improve the performance of a selector device, it is important to understand the characteristics of the switching process. As changes by switching cycle occur at local nanoscale areas, a high-resolution analysis method is needed to investigate this phenomenon. Atomic force microscopy (AFM) is used to analyze the local changes because it offers nanoscale detection with high-resolution capabilities. This review introduces various types of AFM such as conductive AFM (C-AFM), electrostatic force microscopy (EFM), and Kelvin probe force microscopy (KPFM) to study switching behaviors.