http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
- 中文 을 입력하시려면 zhongwen을 입력하시고 space를누르시면됩니다.
- 北京 을 입력하시려면 beijing을 입력하시고 space를 누르시면 됩니다.
RISS 인기검색어
- 검색키워드
- 저자 : Fujinami, Shuhei (검색결과 19 건)
- 무료
- 기관 내 무료
- 유료
-
An Augmented Refrigerator with the Awareness of Wasteful Electricity Usage
Fujinami, Kaori,Kagatsume, Shota,Murata, Satoshi,Alasalmi, Tuomo,Suutala, Jaakko,Roning, Juha The Institute of Internet 2014 International Journal of Internet, Broadcasting an Vol.6 No.1
In this paper, an augmented refrigerator is proposed that presents information to increase the awareness of electric power consumption of a household fridge. The key idea is to reflect wasteful behavior on the feedback to a user, rather than mere amount of consumption or duration of opening a door of a fridge.
-
Deep Learning with Data Augmentation to Add Data Around Classification Boundaries
Hideki Fujinami,Gendo Kumoi,Masayuki Goto 대한산업공학회 2021 Industrial Engineeering & Management Systems Vol.20 No.3
Data augmentation methods are used as a technique to improve generalization by increasing the number of training data in image classification. However, most of these methods are not a data driven algorithm, the degree of improvement of generalization ability by performing these data augmentation methods differs between the domains of image data for training. Generative models are researched to use for augmenting data recently. In particular, Generative Adversarial Networks (GANs) (Goodfellow et al., 2014) that can generate clean image get attention as an excellent innovation in machine learning. As GANs extension method, there is a method called CGANs (Mirza and Osindero, 2014) that can be used for data augmentation. When enough training data for each class are not prepared for classification model, the same is true for training CGANs. In such case, CGAN generates noisy images. This makes a classification model to underfit to the original training data. Moreover, when a CGAN approximates the training data distribution, the CGAN generates new training data in the same region where training data densely exist. In such case, augmented data can’t reduce overfitting on the original training data. Therefore, our research contributes to augment data which meets these two requirements. In this study, we propose a method to generate data by the class specific GAN with small training data and selectively add generated data to the training data set that improves classification accuracy by using the entropy of the classification model. The feature of the proposed method is that it focuses on the positional relationship between data and the classification hyperplane in deep learning. In the proposed method, the entropy of the classification model is used to measure the positional relationship between the classification boundary and the data. As a result, the generalization performance is improved by adding the data around the classification boundary as new training data.
-
-
-
-
-
Science of delicious roods and beautiful women
후지나미츠토무 ( Tsutomu Fujinami ) 한국감성과학회 2019 춘계학술대회 Vol.2019 No.-
We present our experience with craftsmen, where we studied elaborated motions of kneading clay. One can sometimes be tempted to believe that there is one best way to do things well, but our research shows that it depends on one’s inherited characteristics what the best is to do a thing nicely. We then show a result of another research on our taste. We often assume that we taste foods the same as others, but it is not the case. The same sample can be tasted differently among subjects. What he tastes is influenced by what he has eaten and how he lived. We are led to studying typology to deal with the wide ranges of individual differences uniformly by taking a balance between the universal truth and the particular cases. Typology is useful to understand one’s past and the present state, but searching for good plans requires further to incorporate imagination into the design process. The issue is particularly evident when we build a recommendation system for customers. We can induce what he was fond of and can recommend him what to buy, but he may need a change. It forces us to consider another dimension of knowledge to conjecture what he seeks for a change because it cannot be estimated based on the data of his past behaviors. This presentation is concluded with a brief discussion of the methodology to find one’s ideal figure not evident from the data. We studied the motions of craftsmen, who knead the clay as part of preparation for producing ceramics. The task is physically demanding because the clay is heavy and hard. One can somehow master it for a year, but he needs normally five to seven years of practice to truly master the skill. Typically men are stronger than women in strength. This is a primitive typology by which we classify a person to be a male or female although the main factor is strength rather than the gender. Successful male pupils picked up the essence evident in male masters’ motions. An unsuccessful female pupil followed male masters. She could somehow reproduce his motions superficially, but her strength was too weak to produce the same result. She should have learnt from the female master. We turn as the second example to studying tastes. We obtained a taste sensor a few years ago that enables us to evaluate testes of samples as human tongues. We analyzed how consistently subjects are mistaken in judging tastes. We found that what he tastes is largely influenced by what he eats or drinks regularly. One can interpret, for example, a slight hint of sour to be sweet if he drinks too much soft drinks everyday. The liquid usually contains sugar with some refreshing fragrance. The association between sourness and sweetness is so strong in him that he tastes sweet when he is given a slightest hint of sour. One can also taste umami in pure water if he drinks too much coffee everyday. His exposure to bitterness can let him to taste umami in nothing. The phenomenon reminds us of complementary color afterimage in our perception. Typology again plays an important role in implementing the emerging trend of cosmetics. Some distinctive features observed in her appearance are mapped to a particular type such as elegant or boyish. You can start designing your face with the knowledge of your appearance and how you are typed in the map of beautiful women. Finding one’s type of appearance can be automated with the recent technology such as deep learning. Identifying one’s type is however only the half of story. She needs to design her social face with a help of cosmeticians, for which her type only defines the material. The process is parallel with the one in which apprentices of craftsmanship looks for a model to establish his style of kneading the clay. Similarly, she has to curve up her ideal face of the given material. We have seen several cases where individual differences need to be considered as target of science. The science of particulars considers these subtle differences to be important and helps us to enrich our life. A craftsman can produce a nice ceramic. One can eat foods that fit in with his taste. A woman can make herself beautiful. Common to all these cases is that you need to know your type. Truth is relative to type. By pushing the approach to the extreme, we reach to a position in which everyone is unique and the truth is relative to person. We are unsure whether we can still be within the domain of science when we consider each individual as target. We believe that there must be a systematic way to find the best for a person, though. There must be a scientific discipline if we can systematically work on that. Our lifestyle is at least longing for such a support, which is evident for the case of cosmetics.
-
Estimation of fatigue muscle during subjective judgment in muscular exercise
( Atsushi Hori ),( Tsutomu Fujinami ) 한국감성과학회 2019 한국감성과학회 국제학술대회(ICES) Vol.2019 No.-
People who play sports and exercise have been concenred with health. One of the typical exercises is strength training. Desirable training methodologies vary for different purposes of strength training. A training methodology for the purpose of muscle hypertrophy assumes that the effect of exercise is maximized by bringing the active muscles into a state of extreme fatigue. However, it is difficult to properly determine the state of fatigue of the active muscles, and the person cannot help judging it subjectively. We aim to find the characteristics of the biometric data indicating the fatigued state of muscles to determine the state of fatigue. The state is estimated from two types of biological data: electromyogram and muscle temperature. We report the data estimated to be related to muscle fatigue using EMG and muscle temperature.
-
-
EVALUATION OF STRESS REDUCTION BY MENTAL WORK CHARACTERISTIC AND AROMA FRAGRANCE
( Pil Kee Min ),( Tsutomu Fujinami ) 한국감성과학회 2019 한국감성과학회 국제학술대회(ICES) Vol.2019 No.-
People are stressed out as society grows. According to the Comprehensive Survey of Living Conditions(2016), 47.7 percent of people are stressed. We live in today a stress society as the survey suggests. The researches of stress often employ NIRS to observe effects on the brain activity, but it has not been studied how stress can be reduced with aroma fragrances although they must have some effects on brain activity. In this study, we investigate how stress can be reduced by aroma fragrances by examining signals collected from the frontal cortex of subjects using NIRS. We ask twenty participants (male: 5 female:5) to do a quiet task operation (UK test). We do rest-tasking-aromatherapy experiments comparing them. Aroma fragrances are lavender, orange sweet peppermint and lemon
연관 검색어 추천
이 검색어로 많이 본 자료
활용도 높은 자료
