In this paper, we propose a method for no-reference quality assessment of sports videos using 3D shearlet transform and deep residual neural networks. In the proposed method, 3D shearlet transform-based spatiotemporal features are extracted from the overlapped video blocks and applied to logistic regression concatenated with a deep residual neural network based on a conditional video block-wise constraint to learn the spatiotemporal correlation and predict the quality score. Our evaluation reveals that the proposed method predicts the video quality with higher accuracy than the conventional no-reference video quality assessment methods.

An important factor in eye-tracking PC interface for general paralyzed patients is the implementation of the mouse interface, for manipulating the GUI. With a successfully implemented mouse interface, users can generate mouse events exactly at the point of their choosing. However, it is difficult to define this interaction in the eye-tracking interface. This problem has been defined as the Midas touch problem and has been a major focus of eye-tracking research. There have been many attempts to solve this problem using blink, voice input, etc. However, it was not suitable for general paralyzed patients because some of them cannot wink or speak. In this paper, we propose a mouth-pop-up, eye-tracking mouse interface that solves the Midas touch problem as well as becoming a suitable interface for general paralyzed patients using a common RGB camera. The interface presented in this paper implements a mouse interface that detects the opening and closing of the mouth to activate a pop-up menu that the user can select the mouse event. After implementation, a performance experiment was conducted. As a result, we found that the number of malfunctions and the time to perform tasks were reduced compared to the existing method.

Due to the spread of COVID-19, the online lecture has become more prevalent. However, it was found that a lot of students and professors are experiencing lack of communication. This study is therefore designed to improve interactive communication between professors and students in real-time online lectures. To do so, we explore deep learning approaches for automatic recognition of students' facial expressions and classification of their understanding into 3 classes (Understand / Neutral / Not Understand). We use 'BlazeFace' model for face detection and 'ResNet-GRU' model for facial expression recognition (FER). We name this entire process 'Degree of Understanding (DoU)' algorithm. DoU algorithm can analyze a multitude of students collectively and present the result in visualized statistics. To our knowledge, this study has great significance in that this is the first study offers the statistics of understanding in lectures using FER. As a result, the algorithm achieved rapid speed of 0.098sec/frame with high accuracy of 94.3% in CPU environment, demonstrating the potential to be applied to real-time online lectures. DoU Algorithm can be extended to various fields where facial expressions play important roles in communications such as interactions with hearing impaired people.

Segmenting OCT retinal images into layers is important to diagnose and understand the progression of retinal diseases or identify potential symptoms. The task of manually identifying these layers is a difficult task that requires a lot of time and effort even for medical professionals, and therefore, various studies are being conducted to automate this using deep learning technologies. In this paper, we use cGAN-based neural network to automatically segmenting OCT retinal images into seven terrain-type regions defined by six layer boundaries. The network is composed of a Segnet-based generator model and a discriminator model. We also proposed a dynamic programming algorithm for refining the outputs of the network. We performed experiments using public OCT image data set and compared its performance with the Segnet-only version of the network. The experimental results show that the cGAN-based network outperforms Segnet-only version.

The analysis of digital microscopy images plays a vital role in computer-aided diagnosis (CAD) and prognosis. The main purpose of this paper is to develop a machine learning technique to predict the histological grades in prostate biopsy. To perform a multiclass classification, an AI-based deep learning algorithm, a multichannel convolutional neural network (MCCNN) was developed by connecting layers with artificial neurons inspired by the human brain system. The histological grades that were used for the analysis are benign, grade 3, grade 4, and grade 5. The proposed approach aims to classify multiple patterns of images extracted from the whole slide image (WSI) of a prostate biopsy based on the Gleason grading system. The Multichannel Convolution Neural Network (MCCNN) model takes three input channels (Red, Green, and Blue) to extract the computational features from each channel and concatenate them for multiclass classification. Stain normalization was carried out for each histological grade to standardize the intensity and contrast level in the image. The proposed model has been trained, validated, and tested with the histopathological images and has achieved an average accuracy of 96.4%, 94.6%, and 95.1%, respectively.

With the progress of miniaturization and high performance of various sensors, a lot of data is generated in various fields and being collected in real-time, but the use of such large-capacity data is often unable to keep up with the collection technology. In the medical field, health data is collected and managed by platform, which causes inconvenience to users in searching their own health data and receiving medical services. In this paper, in order to solve these problems, we designed and implemented PHDItemReader to improve the speed of data query in a batch application environment that can integrate and process health data having various data expression formats. The experiment compared and analyzed 3 types of query speed based on 1,000,000 hypothetical health data, and as a result of the experiment, it was verified that the PHDItemReader implemented in this paper improved up to about 21% compared to the existing one.

Recently, as the proportion of the elderly population in Korea has increased rapidly, it has entered an aging society. As the scope of social activities of the elderly increases, the risk of safety accidents is also increasing. The traffic accident rate of elderly pedestrians is higher than ordinary people. Although various smart devices for the elderly are researched and developed, electronic components are often used. Electronic devices may not be able to be used if charging is not regularly. In this paper, a smart hat, a smart system for the elderly and the weak, was implemented using solar panels. The system uses solar energy to provide reliable use of smart devices. It is believed to be helpful not only for the elderly but also for the weak people by easy to wear the hat.

As livestock farms grow in size, the number of breeding individuals increases, making it difficult to manage livestock. Livestock farms require an integrated management system such as a monitoring system, an access control system, and an abnormal behavior detection system to manage livestock houses. In this paper, a smart korean cattle livestock management system using IoT and AI technology was proposed for livestock management in livestock farms. The smart korean cattle farm management system consists of a monitoring and control system, a vehicle access management system, and an abnormal cattle behavior detection system. It is expected that this will help manage large-scale livestock houses, and additional research is needed to improve the performance of abnormal behavior detection in the future.

Searching for portable baggage through the system before boarding an airplane at an airport is important because it prevents many risks. In addition to these dangerous items, personal and confidential information leaks are occurring at airports through data storage devices. In the airport search system, there is a need for a system that searches not only dangerous items but also devices that can leak data. In this paper, we proposed a model that searches for a data storage device by improving the existing model. A comparative evaluation was conducted using existing algorithms. As a result, it was confirmed that the performance of the proposed model is 74 in the training data and 46.73 in the test data, which is superior to the existing model.

Aiming at the problem that the existing human behavior recognition algorithm cannot fully utilize the multi-level spatio-temporal information of the network, a human behavior recognition algorithm based on a dense three-dimensional residual network is proposed. First, the proposed algorithm uses a dense block of three-dimensional residuals as the basic module of the network. The module extracts the hierarchical features of human behavior through densely connected convolutional layers; Secondly, the local feature aggregation adaptive method is used to learn the local dense features of human behavior; Then, the residual connection module is applied to promote the flow of feature information and reduced the difficulty of training; Finally, the multi-layer local feature extraction of the network is realized by cascading multiple three-dimensional residual dense blocks, and use the global feature aggregation adaptive method to learn the features of all network layers to realize human behavior recognition. A large number of experimental results on benchmark datasets KTH show that the recognition rate (top-l accuracy) of the proposed algorithm reaches 93.52%. Compared with the three-dimensional convolutional neural network (C3D) algorithm, it has improved by 3.93 percentage points. The proposed algorithm framework has good robustness and transfer learning ability, and can effectively handle a variety of video behavior recognition tasks.

As various styles of fonts were used, there were problems such as output errors due to uninstalled fonts and difficulty in font recognition. To solve these problems, research on font recognition and recommendation were actively conducted. However, Hangul font research remains at the basic level. Therefore, in order to automate the comparison on Hangul font similarity in the future, we analyze the influence of each stroke element property. First, we select seven representative properties based on Hangul stroke shape elements. Second, we design a calculation model to compare similarity between fonts. Third, we analyze the effect of each stroke element through the cosine similarity between the user's evaluation and the results of the model. As a result, there was no significant difference in the individual effect of each representative property. Also, the more accurate similarity comparison was possible when many representative properties were used.

This study examines the game balance of the MOBA game genre, which was selected as a model item for the Asian Games. The "bird-eye view" was used for a more efficient representation of 3D modeling. Based on that, statistical analysis was conducted to present appropriate game camera settings and camera balance to match the competitive structure of the MOBA game. A review of the game camera settings reveals that 64° to 70° is the angle that minimizes the difference in vision between the two-player teams the most. Through a one-way ANOVA analysis, we found that the user ranking level and SVB value are closely related. Therefore, the factor of the regression equation using the SVB value must have a user ranking level. As a result of the optimized camera focus analysis of , the camera setting methods were classified into 3 types. For main action games, the recommended camera angle is 64°~66°, and the recommended camera focus is 11.2 mm~19.3 mm. For action and strategy games, the camera angle is 66°~68°, camera focus - 19.3 mm~27.3 mm. And lastly, for the main strategy game, the recommended camera angle is 68°~70°, and the camera focus is 27.3 mm~35.3 mm.

The purpose of this study is to design and apply YouTube ecosystem using digital curation to improve the interaction of untact class. The untact digital instructional model for building the YouTube ecosystem is composed of four modules: domain, tutor, student, and interface, and the role of each module is described. As an application example, the team project "Development and Operation of Instructional Assistant YouTube" is performed in the Introduction to Media Software class. The learning experience is described in terms of professor, peer tutors, and learners, and learning outcomes are presented through surveys.

This study was conducted on the face ratio of mammals. It can also be applied to character automation by checking factors about the difference between animal and human face shapes. This paper used the face and face area data generated for Deep Learning learning. In detail, the proportion factors of the area comprising the faces of 20 species of animals and humans were defined and the average ratio was calculated. Next, the proportion of each animal was analyzed using the Principal Component Analysis (PCA). Through this, we would like to propose the golden ratio of mammals.