Visual Recognition of Traffic Signs in Natural Scenes Based on Improved RetinaNet

Aiming at recognizing small proportion, blurred and complex traffic sign in natural scenes, a traffic sign detection method based on RetinaNet-NeXt is proposed. First, to ensure the quality of dataset, the data were cleaned and enhanced to denoise. Secondly, a novel backbone network ResNeXt was employed to improve the detection accuracy and effection of RetinaNet. Finally, transfer learning and group normalization were adopted to accelerate our network training. Experimental results show that the precision, recall and mAP of our method, compared with the original RetinaNet, are improved by 9.08%, 9.09% and 7.32%, respectively. Our method can be effectively applied to traffic sign detection.

Development of a Mobile Application to Buy Books through Visual Recognition

Mobile devices have become the most used tool for a large number of tasks that we regularly perform such as relating them, searching for information, and in particular for making purchases. A situation that is frequently repeated in many areas is discovering an object that belongs to another person but we would be interested in being able to acquire it. However, the problem arises of knowing where to buy it. For example, this happens with the clothes that other people are wearing. Today, technology offers recognition mechanisms that can help solve this problem. This article presents an Android app that can recognize a book based on an image and offer places where it can be purchased. For this, Google technology was used to recognize objects from images and it has been combined with the information provided by Google Books to find stores that sell recognized books. In this way, a system has been created that makes it easier for any user to identify and purchase books that they discover at any given time.

Visual recognition for urban traffic data retrieval and analysis in major events using convolutional neural networks

Accurate and prompt traffic data are necessary for the successful management of major events. Computer vision techniques, such as convolutional neural network (CNN) applied on video monitoring data, can provide a cost-efficient and timely alternative to traditional data collection and analysis methods. This paper presents a framework designed to take videos as input and output traffic volume counts and intersection turning patterns. This framework comprises a CNN model and an object tracking algorithm to detect and track vehicles in the camera’s pixel view first. Homographic projection then maps vehicle spatial-temporal information (including unique ID, location, and timestamp) onto an orthogonal real-scale map, from which the traffic counts and turns are computed. Several video data are manually labeled and compared with the framework output. The following results show a robust traffic volume count accuracy up to 96.91%. Moreover, this work investigates the performance influencing factors including lighting condition (over a 24-h-period), pixel size, and camera angle. Based on the analysis, it is suggested to place cameras such that detection pixel size is above 2343 and the view angle is below 22°, for more accurate counts. Next, previous and current traffic reports after Texas A&M home football games are compared with the framework output. Results suggest that the proposed framework is able to reproduce traffic volume change trends for different traffic directions. Lastly, this work also contributes a new intersection turning pattern, i.e., counts for each ingress-egress edge pair, with its optimization technique which result in an accuracy between 43% and 72%.

Implementasi Augmented Reality pada Mata Pelajaran Komputer dan Jaringan Dasar di SMK Nufa Citra Mandiri Berbasis Android

In the process of teaching and learning activities in schools, there are still many that only use learning media in the form of books. This causes boredom in the learning process. Especially in basic computer and network subjects, where in this subject there is a lot of visual recognition and developing forms of technology. Good visual presentation is very important to support the learning process. Augmented Reality (AR) is a learning medium that is currently still being used in the teaching and learning process. AR is an application that combines the real world with the virtual world in two-dimensional or three-dimensional forms which are projected in a real environment at the same time. AR technology is very appropriate to be implemented to support the visualization process in learning activities, because it can present visualization in 3-dimensional form. So that the visualization is more interactive and more interesting. In the learning process at SMK Nufa Citra Mandir, they still rely on books only, especially in computer and basic network subjects which really need visualization of the devices discussed. The application of AR technology to basic computer and network subjects at SMK Nufa Citra Mandiri is expected to help in the learning process. In its development, the design method used is the Extreme Programming (XP). Purpose of this research is to create software with AR technology on Android platform, so that students can better understand and get to know computer tools and computer network devices contained in basic computer and network subjects.

A Sensor-Based Drone for Pollutants Detection in Eco-Friendly Cities: Hardware Design and Data Analysis Application

The increase in produced waste is a symptom of inefficient resources usage, which should be better exploited as a resource for energy and materials. The air pollution generated by waste causes impacts felt by a large part of the population living in and around the main urban areas. This paper presents a mobile sensor node for monitoring air and noise pollution; indeed, the developed system is installed on an RC drone, quickly monitoring large areas. It relies on a Raspberry Pi Zero W board and a wide set of sensors (i.e., NO2, CO, NH3, CO2, VOCs, PM2.5, and PM10) to sample the environmental parameter at regular time intervals. A proper classification algorithm was developed to quantify the traffic level from the noise level (NL) acquired by the onboard microphone. Additionally, the drone is equipped with a camera and implements a visual recognition algorithm (Fast R-CNN) to detect waste fires and mark them by a GPS receiver. Furthermore, the firmware for managing the sensing unit operation was developed, as well as the power supply section. In particular, the node’s consumption was analysed in two use cases, and the battery capacity needed to power the designed device was sized. The onfield tests demonstrated the proper operation of the developed monitoring system. Finally, a cloud application was developed to remotely monitor the information acquired by the sensor-based drone and upload them on a remote database.

Development of a Standardised Process of Scenario Writing Through a Virtual Assistant Model in Nursing Teaching

Background: The use of simulation allows students to develop skills. Simulation in the teaching of clinical skills is preceded by some stages, the first of which requires the development of a scenario. This study aimed to develop and test a virtual model for creating scenarios for realistic simulation, focusing on the person with dependence in self-care activities. Methods: A methodological study was conducted in two phases. The first phase of the study aimed to analyse and propose the structure and functioning of the virtual assistant for scenario creation through the nominal group’s technique, involving a group of 10 experts. The second, a quasi-experimental study without a control group, with 128 second-year students, in the four-year nursing degree course, who participated in two moments of realistic simulation, one with a traditional scenario and the other with a scenario built through the virtual assistant. The students completed a questionnaire to assess their understanding of the data, suggested interventions, and their contribution to learning after each simulation experience. Results: The group of experts identified the fields and key concepts that should be part of the structure of the scenarios and proposed a set of icons for better visual recognition of the information. Students considered that the new scenario template favoured their understanding of the situation under analysis and the recognition of the focuses of attention that they should prioritise for the elaboration of the intervention plan. Conclusions: A virtual role-play assistant model for a standardized process of scenario writing to help realistic simulation in nursing teaching is a novelty in this study likely to contribute to learning gains.

The Influence of Negative Orthographic Neighborhood in the Lexical Decision Task: Valence and Arousal Contributions

This study investigated whether the visual recognition of neutral words might be influenced by the emotional dimensions (i.e., valence and arousal) of orthographically similar lexical representations, and whether this might also depend on emotional-related traits of participants (i.e., alexithymia). To this end, 108 participants performed a lexical decision task with 80 neutral words with a higher frequency orthographic neighbor that varied in valence (from neutral to negative) and arousal (from low to high). The main finding was the expected interaction effect between the valence and arousal of the neighbor on the lexical decision times of neutral stimulus words. Longer reaction times were found when the valence score of the neighbor decreased from neutral to negative for words with a low-arousal orthographic neighbor while this emotional neighbor effect was reversed for words with a high-arousal negative neighbor. This combined influence of the valence and arousal of the neighbor was interpreted in terms of increased lexical competition processes and direct influence of the affective system on the participant’s response. Moreover, this interaction effect was smaller when the level of alexithymia of the participants increased, suggesting that people with a higher level of alexithymia are less sensitive to the emotional content of the neighbor. The results are discussed within an interactive activation model of visual word recognition incorporating an affective system with valence and arousal dimensions, with regard to the role of the alexithymia level of participants.

Probing the Effect of Selection Bias on Generalization: A Thought Experiment

Learned systems in the domain of visual recognition and cognition impress in part because even though they are trained with datasets many orders of magnitude smaller than the full population of possible images, they exhibit sufficient generalization to be applicable to new and previously unseen data. Since training data sets typically represent such a small sampling of any domain, the possibility of bias in their composition is very real. But what are the limits of generalization given such bias, and up to what point might it be sufficient for a real problem task? Although many have examined issues regarding generalization from several perspectives, this question may require examining the data itself. Here, we focus on the characteristics of the training data that may play a role. Other disciplines have grappled with these problems also, most interestingly epidemiology, where experimental bias is a critical concern. The range and nature of data biases seen clinically are really quite relatable to learned vision systems. One obvious way to deal with bias is to ensure a large enough training set, but this might be infeasible for many domains. Another approach might be to perform a statistical analysis of the actual training set, to determine if all aspects of the domain are fairly captured. This too is difficult, in part because the full set of important variables might not be known, or perhaps not even knowable. Here, we try a different, simpler, approach in the tradition of the Thought Experiment, whose most famous instance is perhaps Schrödinger's Cat, to address part of these problems. There are many types of bias as will be seen, but we focus only on one, selection bias. The point of the thought experiment is not to demonstrate problems with all learned systems. Rather, this might be a simple theoretical tool to probe into bias during data collection to highlight deficiencies that might then deserve extra attention either in data collection or system development.