Abstract This paper describes the world’s largest gait database with wide view variation, “OU-ISIR database, multi-view large population dataset (OU-MVLP)”, and its application to a statistically reliable performance evaluation of vision-based cross-view recognition. Specifically, we construct that includes 10,307 subjects (5114 males 5193 females) from 14 angles ranging 0° −90°, 180° −270°. In addition, evaluate various approaches recognition which are robust against angles. By using our...

In this paper, we discuss input/output architectures for convolutional neural network (CNN)-based cross-view gait recognition. For purpose, consider two aspects: verification versus identification and the tradeoff between spatial displacements caused by subject difference view difference. More specifically, use Siamese with a pair of inputs contrastive loss triplet ranking identification. The aforementioned CNN are insensitive to displacement, because matching is calculated at last layer...

Abstract In this paper, we describe the world’s largest gait database with real-life carried objects (COs), which has been made publicly available for research purposes, and its application to performance evaluation of vision-based recognition. Whereas existing databases recognition include at most 4007 subjects, constructed an extremely large-scale that includes 62,528 equal distribution males females, ages ranging from 2 95 years old. Moreover, whereas consider a few predefined CO...

Abstract Gait-based age estimation has been extensively studied for various applications because of its high practicality. In this paper, we propose a gait-based method using convolutional neural networks (CNNs). Because gait features vary depending on subject’s attributes, i.e., gender and generation, the following three CNN stages: (1) estimation, (2) age-group (3) regression. We conducted experiments large population database confirm that proposed outperforms state-of-the-art benchmarks.

Abstract Gait-based features provide the potential for a subject to be recognized even from low-resolution image sequence, and they can captured at distance without subject’s cooperation. Person recognition using gait-based (gait recognition) is promising real-life application. However, several body parts of subjects are often occluded because beams, pillars, cars trees, or another walking person. Therefore, not applicable approaches that require an unoccluded gait sequence. Occlusion...

Tomoyuki Kajiwara, Chenhui Chu, Noriko Takemura, Yuta Nakashima, Hajime Nagahara. Proceedings of the 2021 Conference North American Chapter Association for Computational Linguistics: Human Language Technologies. 2021.

Climate change is one of the most important issues for humanity. To defuse this problem, it considered necessary to improve energy efficiency, make sources cleaner, and reduce consumption in urban areas. The Japanese government has recommended an air conditioner setting 28 ℃ summer 20 winter since 2005. aim save by keeping room temperatures constant. However, unclear whether appropriate temperature workers students. This study examined thermal environments influence task performance over...

Gait-based age estimation is a key technique for many applications. It well known that uncertainty highly dependent on (i.e., small children and large adults), it important to know the above-mentioned Therefore, we propose method uncertainty-aware gait-based by introducing label distribution learning framework. Specifically, design network takes an appearance-based gait feature as input outputs discrete distributions in integer domain. We then train minimize loss function, which defined...

Dynamic facial expression recognition (DFER) is an important task in the field of computer vision. To apply automatic DFER practice, it necessary to accurately recognize ambiguous expressions, which often appear data wild. In this paper, we propose MIDAS, a augmentation method for DFER, augments with soft labels consisting probabilities multiple emotion classes. training are augmented by convexly combining pairs video frames and their corresponding class labels, can also be regarded as...

Learners are expected to stay wakeful and focused while interacting with e-learning platforms. Although wakefulness of learners strongly relates educational outcomes, detecting drowsy learning behaviors only from log data is not an easy task. In this study, we describe the results our research model learners' based on multimodal generated heart rate, seat pressure, face recognition. We collected in a blended course informatics conducted two types analysis them. First, clustered features...

This paper describes a deep learning approach to classify physically fatigued and non-fatigued gait cycles via recurrent neural network (RNN), where each cycle is represented as time series of three-dimensional coordinates body joints. Gait inherently have large intra-class variations caused by stance differences (e.g., which foot supporting/swinging) at the beginning cycle, makes it difficult identify subtle induced fatigue. To overcome these difficulties, we introduce supporting foot-aware...

This study aimed to clarify the adaptation features of University students exposed fully online education during novel coronavirus disease 2019 (COVID-19) pandemic and identify accompanying mental health problems predictors school adaptation. The has forced many universities transition rapidly delivering education. However, little is known about impact this drastic change on students' cross-sectional used an questionnaire, including assessments impressions education, engagement, health,...

This paper describes a view planning of multiple cameras for tracking persons surveillance purposes. When only few active are used to cover wide area, their views is an important issue in realizing competent system. We develop multi-start local search (MLS)-based method which iteratively selects fixation points the by expected number tracked maximized. Considering fact that person's motion can be estimated with its intermittent observations, we set criterion encourages frequent shifts and...

Deep convolutional neural networks (CNNs) have established their feet in the ground of computer vision and machine learning, used various applications. In this work, an attempt is made to learn a CNN for task facial expression recognition (FER). Our network has layers linked with FC layer skip-connection classification layer. Motivation behind design that lower are responsible level features, expressions can be mainly encoded low-to-mid features. Hence, order leverage responses from layers,...

Drowsiness is a major factor that hinders learning. To improve learning efficiency, it important to understand students' physical status such as wakefulness during online coursework. In this study, we have proposed drowsiness estimation method based on learners' head and facial movements while viewing video lectures. examine the effectiveness of in estimation, collected learner data recorded e-learning applied deep approach under following conditions: (a) using only movement data, (b) (c)...

Gait-based age estimation is one of key techniques for many applications (e.g., finding lost children/aged wanders). It well known that the uncertainty highly dependent on ages (i.e., it generally small children while large adults/the elderly), and important to know above-mentioned applications. We therefore propose a method uncertainty-aware gait-based by introducing label distribution learning framework. More specifically, we design network which takes an appearance-based gait feature as...

This paper deals with a multiagent path-planning problem where several robots track humans to obtain detailed information on human behaviors and characteristics. For this, agents' paths are planned the basis of similarity between predicted positions field view. The long-horizon path an accurate prediction improves tracking performance. However, it requires heavy computation is less useful if inaccurate. Since accuracy depends situation, term determined by current previous predictions....

Abstract Topic modeling that can automatically assign topics to legal documents is very important in the domain of computational law. The relevance modeled strongly depends on context they are used in. On other hand, references laws and prior cases key elements for judges rule a case. Taken together, these form network, whose structure be analysed with network analysis. However, content referenced may not always accessed. Even case, reference itself shows share latent similar...