Accepted Papers


SURFACE CRACK DETECTION USING HIERARCHAL CONVOLUTIONAL NEURAL NETWORKS

Davis Bonsu Agyemang and Mohamed Bader-El-Den
The School of Computing, University of Portsmouth, Buckingham Building, Lion Terrace, Portsmouth PO1 3HE.

ABSTRACT

Cracks on concrete walls can imply that a building possesses issues with its structural integrity. Surveyors who inspect these defects are expected to provide their customers with excellent evaluation regarding its severity. The process is currently conducted through visual inspection, resulting in occasions of subjective judgements being made on the classification and severity of the concrete crack which poses danger for customers and the environment as it not being analysed objectively. Many researchers have applied numerous classification techniques to tackle this issue but from the author's knowledge, their methods do not provide the severity levels of concrete cracks and have no feedback mechanism for adaptability of when their method classifies incorrectly. In this paper, the author proposes in building a mobile application with the 2 mentioned capabilities and using a trained Hierarchal-Convolutional Neural Network(H-CNN) to evaluate the concrete surface via images taken via the mobile device.

KEYWORDS

Concrete cracks,classification algorithms, Convolutional Neural Networks(CNN) ,Hierarchical Convolutional Neural Network(H-CNN), accuracy,loss,Python,Flask, HTM,CSS &Javascript.


UVAS-BASED SMALL OBJECT DETECTION AND TRACKING IN VARIOUS COMPLEX SCENARIOS

Shicheng Zu, Yanwen Hu, Zhongzheng Yu, Wanqun Chen, and Kai Yang , National Defense Science and Technology Coordination and Innovation Base, Nanjing, China

ABSTRACT

We have witnessed drastic progress in object detection in recent years due to the blooming development of neural networks. Most mainstream object detectors are sensitive to objects of regular scale because their detection largely depends on deep convolutional feature maps lacking low-level features, i.e. edges, and contours, which contribute to small object detection. Our study focuses on UVAs-based small object detection at a high altitude, i.e. 100 meters. We construct a pipeline by integrating the moving foreground segmentation algorithm, the conventional CNN model, the boosted cascaded classifier, and the tracker that can detect and track the small object progressively in a cascaded manner. We perform the qualitative and quantitative evaluation of the detection and tracking performance of our constructed pipeline in various complex conditions. The comparison study confirms its superiority in small object detection and strong robustness against various influential nuisances. Based on our constructed pipeline, we develop a real-time UVAs-based small object detection and tracking system. The systematic architecture and the general steps taken by the UVAs to realize small object detection are also presented. Finally, we qualitatively and quantitatively evaluate 8 popular trackers based on the attributes of the image sequences. The most appropriate tracker can be determined in response to a specific condition. Our study testifies that by taking advantage of the merits of each algorithm germane to a given task, the performance can be further improved

KEYWORDS

UVAs, Small Object Detection, Foreground Segmentation Algorithm, CNN, Tracker


ABNORMALITY DETECTION IN MUSCULOSKELETAL RADIOGRAPHS BY DENSENET AND INCEPTION-V3

Zahra Rezaei1, Hossein Ebrahimpour Komleh2, Behnaz Eslami3, Kaveh Daneshmand Jahromi4, Ramyar Chavoshinejad5 1, 2Department of Computer and Electrical Engineering University of Kashan, Kashan, Iran, 3Department of Computer Engineering, Science and Research Branch Islamic Azad University, Tehran, Iran, 4Post-MBA students in Business Intelligence (BI), Industrial Management Institute (IMI), Tehran, Iran and 5Mabna Veterinary Labs, Karaj, Alborz, Iran

ABSTRACT

One of the most remarkable applications of deep learning is in medical diagnoses and new improvements in this field have shown that with large enough datasets and right methods, one can achieve results as reliable as experienced doctors. One of such developments is MURA which is a dataset about musculoskeletal radiographs consisting of 14,863 studies from 12,173 patients, resulting in a number of 40,561 multi-view radiographic images. Each one of these studies is about one of seven standard upper extremity radiographic study types, namely, finger, forearm, elbow, hand, shoulder, homeruns, and wrist. Each study was categorized as normal or abnormal by board-certified radiologists in the diagnostic radiology environment between 2001 and 2012. Abnormality detection in muscular radiography is of great clinical applications. This gains more importance in cases which abnormality detection is difficult for physicians. If the proposed model can help us in detection, the process of treatment will precipitate. This model is termed inception-v3. In this study, we evaluated the MURA data set through Dense NET and inception-v3 methods. The results indicated that the former had better performance, and we added pre-processing module to it in order to improve the accuracy of the DenseNet method to detect abnormality. In this context, we train machine to be sensitive to presence of external objects to be distinguished with actual abnormality like bone fraction; we achieved that by a lot of various radiographs as machine inputs. By this strategy, both techniques (DenseNet and Inception-V3) showed improvement in accuracy. Thus, we subgroup abnormality into with or without the presence of external objects. Although the average opinion of radiologists still shows better results, in images in which fracture detection is delicate, like finger fracture, the proposed model works more accurately, and it can as a decision support assistant for physicians in final detection of fracture. If the image is separated from normal images using Platinum, and a new class is made, and pre-processing is done, the precision of the proposed model enhances. So, a model which can automatically detect abnormality can identify the part of image which is detected to be abnormal by the model. If this model is efficient, it can interpret the images more efficiently, it can reduce errors, and it can enhance quality. In order to evaluate the integration of this model with other models of deep learning in clinical setting, more studies are needed to be carried out

KEYWORDS

Musculoskeletal radiographs, deep learning, medical image processing, abnormal detection


Application Development using gamification techniques to simulate the admission exam to University

Michael Valencia Ramírez, Universidad Distrital Francisco José de Caldas Bogotá, Colombia

ABSTRACT

This article aims to show the advances in the development of a mobile application to simulate the admission exam to the National University of Colombia (UN), which makes use of gamification techniques. For young graduates of high school, one of the main concerns is to apply to the university and be admitted. The National University of Colombia, because it is public in nature and has a high academic level, it is in great demand in the country, each semester there are approximately 72 thousand applicants and only about 10% is accepted due to the low availability of university quotas. The selection process is carried out by means of an admission examination, which evaluates the skills in reading comprehension and logical-mathematical reasoning of the aspirant. Faced with this problem and with the unstoppable evolution of technology, it is vital to integrate the ICT tools (Information and Communication Technologies), as components in learning and preparing them, a mobile application is developed, with the aim of helping aspirants to have a better preparation for the exam. It is taken as population study of 20 students of the eleventh grade belonging to the Liceo Cultural Ernesto Guhl School in Bogota, Colombia.

KEYWORDS

Education, Gamification, University, ICT, Mobile application


Mobile application focused on knowledge training intended for “Saber 11 Test” using Augmented Reality Environments

Maily A. Quintero E Universidad Distrital Francisco José de Caldas Bogotá, Colombia

ABSTRACT

In Colombia, the system of evaluation and promotion of education is carried out by the ICFES through state tests applied to grades 3rd, 5th, 7th, 9th and 11th. The Saber 11 Test is an assessment of the basic competences that is performed on grade 11th students. This test scores are used as an important requirement for admission to college. In 2018, 37% of students scored low and failed to apply to the multiple benefits offered by government and universities. This is because the current training mechanisms do not generate impact and motivation in the student so that he can take advantage of the training material offered by various instances. The development of a mobile application prototype for devices based on Android platform is presented, using Augmented Reality tools, which adds virtual elements to the real environment while providing information of interest to the user and takes advantage of the infrastructure of ICT (Information and Communication Technologies). Being a motivational learning strategy helping the student to become familiar with the structure, areas and skills, and proposed questions on the exam know 11.

KEYWORDS

Education, Augmented Reality, Mobile Application, Technology, Saber 11


SURFACE CRACK DETECTION USING HIERARCHAL CONVOLUTIONAL NEURAL NETWORKS

Davis Bonsu Agyemang and Mohamed Bader-El-Den
The School of Computing, University of Portsmouth, Buckingham Building, Lion Terrace, Portsmouth PO1 3HE.

ABSTRACT

Cracks on concrete walls can imply that a building possesses issues with its structural integrity. Surveyors who inspect these defects are expected to provide their customers with excellent evaluation regarding its severity. The process is currently conducted through visual inspection, resulting in occasions of subjective judgements being made on the classification and severity of the concrete crack which poses danger for customers and the environment as it not being analysed objectively. Many researchers have applied numerous classification techniques to tackle this issue but from the author's knowledge, their methods do not provide the severity levels of concrete cracks and have no feedback mechanism for adaptability of when their method classifies incorrectly. In this paper, the author proposes in building a mobile application with the 2 mentioned capabilities and using a trained Hierarchal-Convolutional Neural Network(H-CNN) to evaluate the concrete surface via images taken via the mobile device.

KEYWORDS

Concrete cracks,classification algorithms, Convolutional Neural Networks(CNN) ,Hierarchical Convolutional Neural Network(H-CNN), accuracy,loss,Python, Flask, HTM,CSS &Javascript.


AUTOMATIC DESIGN SPACE EXPLORATION OF HETEROGENEOUS EMBEDDED SYSTEMS USING SPECIALIZED GENETIC ALGORITHM

Mouna HALIMA, Lobna Kriaa and Leila Azzouz Saidane, ENSI University, Manouba, Tunisia

ABSTRACT

Design space exploration (DSE) techniques have become essential in order to find the best compromise between different design goals and their trade-off. Designers need effective optimization techniques to support their design decisions. However, the partitioning problem is one of the most important issues.Many heuristic methods have been proposed to solve this problem. This paper introduces an automatic design space exploration approach which is based on a customized Genetic Algorithm, named Binary Genetic Algorithm (BGA). This new approach helps to find rapidly the best solution satisfying a set of systems requirements and/or constraints. Different steps, from the system specification to the target architecture including mapping and scheduling objectives, compose our flow.

KEYWORDS

Codesign embedded system, DSE, architecture exploration, partitioning, optimization algorithms,Genetic Algorithm


THE IMPACT OF QUANTUM GENETIC ALGORITHMS IN MINIMIZING TASK MIGRATION’ OVERHEADS

Belkebir Djalila, Department of Computer Engineering, Kasdi Merbah Ouargla University, Ouargla, Algeria

ABSTRACT

Genetic algorithms are widely used to solve problems in complex system for its easiest implementation and achievability of global optimum with a proper approximate solution, but quantum-inspired evolutionary algorithms have been surpassing the classical algorithms for their abilities to solve problems of polynomial-time that is considered it as impossible to be solved but with million years while their advantage is to balance between exploration and exploitation of the solution space and also obtain better solutions, even with a small population. In this paper, we propose a novel task migration algorithm that minimize the overall overheads caused during the migration process in network on chip (NoC) while searching the appropriate checkpoints during run-time using quantum genetic algorithm.

KEYWORDS

Genetic Algorithms, Quantum Computing, Mapping, Scheduling, Network on Chip, Task Migration