Alternatives for Locating People Using Cameras and Embedded AI Accelerators: A Practical Approach

Indoor positioning systems usually rely on RF-based devices that should be carried by the targets, which is non-viable in certain use cases. Recent advances in AI have increased the reliability of person detection in images, thus, enabling the use of surveillance cameras to perform person localization and tracking. This paper evaluates the performance of indoor person location using cameras and edge devices with AI accelerators. We describe the video processing performed in each edge device, including the selected AI models and the post-processing of their outputs to obtain the positions of the detected persons and allow their tracking. The person location is based on pose estimation models as they provide better results than do object detection networks in occlusion situations. Experimental results are obtained with public datasets to show the feasibility of the solution.

A Fast and Accurate System for Face Mask Detection in Public Places

Changes in the way of life of everybody around the world. In those progressions wearing a cover has been indispensable to each person. Location of individuals who are not wearing veils is a test because of Outbreak of the Coronavirus pandemic has made different the enormous number of populaces. The COVID-19 pandemic has reshaped life. Numerous of us are remaining at home, staying away from individuals in the city and changing every day propensities, such as going to class or work, in manners we never envisioned. While we are changing old practices, there are new schedules we need to receive. Most importantly is the propensity for wearing a veil or face covering at whatever point we are in a public space. Veils and face covers can forestall the wearer from communicating the COVID-19 infection to other people and may give some assurance to the wearer. All-inclusive cover use can altogether diminish infection transmission locally by forestalling anybody, including the individuals who are accidentally conveying the infection, from communicating it to other people. Along these lines, the significance of wearing veil and its identification is exceptionally clear. Face veil recognition frameworks are currently progressively significant, particularly in keen medical clinics for viable patient consideration. They're likewise significant in arenas, air terminals, stockrooms, and other swarmed spaces where pedestrian activity is weighty and security guidelines are basic to defending everybody's wellbeing. Face veil recognition framework can guarantee our security and the security of others. This task can be utilized in schools, clinics, banks, air terminals, and so forth as a digitized examining instrument. The procedure of recognizing individuals' countenances and isolating them into two classes in particular individuals with covers and individuals without covers is finished with the assistance of picture handling and profound learning. With the assistance of this task, an individual who is expected to screen individuals can be situated in a far off region furthermore, still can screen productively and give directions appropriately. Different libraries of python like Open CV, Tensor Flow and Keras are utilized. In Deep Learning Convolution Neural Networks is a class Deep Neural Networks which is used to prepare the models in this task.

Rule-based Named Entity Recognition (NER) to Determine Time Expression for Balinese Text Document

Named Entity Recognition (NER) is a process to identify words or phrases as a named entity, such as a person, location, time expression, or organization. In this research, we are interested in developing a NER which able to identify the time expression entity in Balinese text documents. The time expression entity becomes an important component in the text because it is usually followed by important facts and information. NER was built using a rules-based approach. The rules are built based on direct observation of documents and pay attention to the morphological and contextual structures. Based on the experiments conducted, the average results of the precision, recall, and f-measure values were 0.85, 0.87, and 0.85.

Comparing the use of sum-scores and Item Response Theory’s person location scores as measures of Functional Somatic Symptoms

Purpose: This study aims to compare the use of sum-scores and person location scores from Item Response Theory (IRT) as outcome measures of Functional Somatic Symptoms (FSS) in an epidemiological study. Method: Data from 1247 participants (60% female) from the Tracking Adolescents' Individual Lives Survey (TRAILS) general population cohort study at the fifth (mean age = 22.2, SD = 0.64) and sixth (mean age = 25.6, SD = 0.6) measurement waves was employed. We fitted the Graded Response Model (GRM) from IRT to the 12 items of the “physical complaints” subscale of the Adult Self-Report (ASR) to calculate item and person location parameters. We performed bootstrapped multiple linear regressions to analyze the relationship between Positive Affect (PA) and FSS using person location scores and compared the results to results obtained using sum-scores. Results: The items “nausea” and “abdominal pain” were most discriminative. ASR sum-scores and person location scores were highly correlated, although the latter captured more variability. Using sum-scores and person location scores to study the association between PA and FSS did not result in relevant differences. Conclusion: Although person location scores capture more variability, we did not find added value in the longitudinal analyses of the association between PA and FSS.

A French corpus annotated for multiword expressions and named entities

We present the enrichment of a French treebank of various genres with a new annotation layer for multiword expressions (MWEs) and named entities (NEs).1 Our contribution with respect to previous work on NE and MWE annotation is the particular care taken to use formal criteria, organized into decision flowcharts, shedding some light on the interactions between NEs and MWEs. Moreover, in order to cope with the well-known difficulty to draw a clear-cut frontier between compositional expressions and MWEs, we chose to use sufficient criteria only. As a result, annotated MWEs satisfy a varying number of sufficient criteria, accounting for the scalar nature of the MWE status. In addition to the span of the elements, annotation includes the subcategory of NEs (e.g., person, location) and one matching sufficient criterion for non-verbal MWEs (e.g., lexical substitution). The 3,099 sentences of the treebank were double-annotated and adjudicated, and we paid attention to cross-type consistency and compatibility with thesyntactic layer. Overall inter-annotator agreement on non-verbal MWEs and NEs reached 71.1%. The released corpus contains 3,112 annotated NEs and 3,440 MWEs, and is distributed under an open license.

Make or break it: boundary conditions for integrating multiple elements in episodic memory

Event memories are characterized by the holistic retrieval of their constituent elements. Studies show that memory for individual event elements (e.g. person, object and location) are statistically related to each other, and that the same associative memory structure can be formed by learning all pairwise associations across separated encoding contexts (person–object, person–location, object–location). Counter to previous studies that have shown no differences in holistic retrieval between simultaneously and separately encoded event elements, adults did not show evidence of holistic retrieval from separately encoded event elements when using a similar paradigm adapted for children (Experiment 1). We conducted a further five online experiments to explore the conditions under which holistic retrieval emerges following separated encoding of within-event associations, testing for influences of trial length (Experiment 2), the number of events learned (Experiment 3a) and stimulus presentation format (Experiments 3b, 4a, 4b). Presentation of written words was optimal for integrating elements across encoding trials, whereas the addition of spoken words disrupted integration across separately presented associations. The use of picture stimuli also produced effect sizes smaller than those of previously published research. We discuss the ways in which memory integration processes may be disrupted by these differences in presentation format. The findings have practical implications for the utility of this paradigm across research and learning contexts.

CustNER

This article describes CustNER: a system for named-entity recognition (NER) of person, location, and organization. Realizing the incorrect annotations of existing NER, four categories of false negatives have been identified. The NEs not annotated contain nationalities, have corresponding resource in DBpedia, are acronyms of other NEs. A rule-based system, CustNER, has been proposed that utilizes existing NERs and DBpedia knowledge base. CustNER has been trained on the open knowledge extraction (OKE) challenge 2017 dataset and evaluated on OKE and CoNLL03 (Conference on Natural Language Learning) datasets. The OKE dataset has also been annotated with the three types. Evaluation results show that CustNER outperforms existing NERs with F score 12.4% better than Stanford NER and 3.1% better than Illinois NER. On another standard evaluation dataset for which the system is not trained, the CoNLL03 dataset, CustNER gives results comparable to existing systems with F score 3.9% better than Stanford NER, though Illinois NER F score is 1.3% better than CustNER.

Make or break it: Boundary conditions for integrating multiple elements in episodic memory

Event memories are characterised by the holistic retrieval of their constituent elements. Studies show that memory for individual event elements (e.g., person, object, and location) are statistically related to each other, and that the same associative memory structure can be formed by learning all pairwise associations across separated encoding contexts (person-object, person-location, object-location). Counter to previous studies that have shown no differences in holistic retrieval between simultaneously and separately encoded event elements, adults did not show evidence of holistic retrieval from separately encoded event elements when using a similar paradigm adapted for children (Experiment 1). We conducted a further five online experiments to explore the conditions under which holistic retrieval emerges following separated encoding of within-event associations, testing for influences of trial length (Experiment 2), the number of events learned (Experiment 3a), and stimulus presentation format (Experiments 3b, 4a, 4b). Presentation of written words was optimal for integrating elements across encoding trials, whereas the addition of spoken words disrupted integration across separately presented associations. Use of picture stimuli also produced effect sizes smaller than those of previously published research. We discuss the ways in which memory integration processes may be disrupted by these differences in presentation format. The findings have practical implications for the utility of this paradigm across research and learning contexts.