論文 - 馬場口　登

Formulating what a deceived person thinks 査読

S. Myojin, N. Babaguchi

23rd International Symposium on Artificial Life and Robotics, AROB 23rd 2018   2018年1月

Development of a Stroll Support System Using Route Display on a Map and Photograph Sharing Service. 査読

J. Itou, T. Mori, J. Munemori, N. Babaguchi

Proc. of 10th International Conference on Collaboration Technologies (CollabTech2018)   48 - 55   2018年

Constructing Geospatio-Temporal Concept Graphs from Tagged Images 査読

H. Honjo, N. Nitta, K. Nakamura, N. Babaguchi

Proceedings - 2017 IEEE 3rd International Conference on Multimedia Big Data, BigMM 2017   169 - 176   2017年6月

　詳細を見る

This paper proposes a method for discovering the knowledge about geospatio-temporal concepts, which are strongly related to specific locations and times, from images tagged with latitude and longitude coordinates, time-stamps, and textual tags. The proposed method firstly extracts textural tags used either in specific locations or times as the geospatial or temporal concepts with their location or time information. Then, the semantic relations among the geospatio-temporal concepts are extracted based on the co-occurrence of the corresponding tags to be represented as a graph. Finally, visually discriminative geospatio-temporal concepts are linked to representative visual concepts. The usefulness of the discovered knowledge is evaluated based on the automatic tagging for the images tagged with latitude and longitude coordinates and time-stamps.

Extracting Real-World Observations from Microblog 査読

M. Yoshitake, Naoko Nitta, Kazuaki Nakamura, Noboru Babaguchi

Proceedings - 2017 IEEE 3rd International Conference on Multimedia Big Data, BigMM 2017   232 - 237   2017年6月

　詳細を見る

Since a large number of users of various social networking services post what they observe around themselves, what is going on around the world can be known in real time by extracting such real-world observations. Especially, observations covering miscellaneous areas of interest are posted to Twitter as short text messages. Our goal is to extract such observations to better understand the current situations of the real world. Since the observations at specific locations are often described with words representing the observed locations or events, the locations or events all over the world can be discovered by finding such local words which are uniquely used at specific locations or times. Then, parts of their observations can be obtained by using the local words to define the semantics of each local word. Finally, the observations which do not contain the local words but are related to the locations or events can be extracted based on their semantic relevancy to each local word.

On-line Geospatial Term Extraction from Streaming Geotagged Tweets 査読

T. Kamimura, N. Nitta, K. Nakamura, N. Babaguchi

Proceedings - 2017 IEEE 3rd International Conference on Multimedia Big Data, BigMM 2017   322 - 329   2017年6月

　詳細を見る

Recently, geotagged posts to social media such as Twitter have become major sources for geospatial information referenced to a set of geographic coordinates. Especially, many existing work collects geospatial terms which identify geographic locations by examining the spatial locality of the term usage patterns observed in the geotagged posts accumulated for a certain period of time. Although the spatial locality of the collected geospatial terms can only be temporary, such time variability difference among the geospatial terms have not been considered in the existing work. Thus, in order to separately collect the stationary and temporary geospatial terms with proper timing, we propose an on-line method for constructing a geographical dictionary containing the up-to-date geospatial terms and their locations by continuously examining the spatial locality of terms in streaming geotagged posts. The geospatial terms can be distinguished between stationary and temporary terms according to their usage patterns in the recent set of geotagged posts and their locations recorded in the geographical dictionary. Additionally, images representing each geospatial term can also be collected from the posts containing the corresponding term. The usefulness of the collected geospatial information is demonstrated in comparison to existing geographical dictionaries constructed by experts, crowdsourcing, and batch method.

［解説］グループごとの行動状態を考慮した群衆映像からのグループ検出 招待

中村和晃, 馬場口登

画像ラボ   28 ( 5 )   1 - 10   2017年5月

Detection of groups in crowd considering their activity state 査読

K. Nakamura, T. Ono, N. Babaguchi

Proceedings - International Conference on Pattern Recognition   277 - 282   2017年4月

　詳細を見る

In this paper, we focus on the problem of group detection in crowd, which is a task of partitioning a set of pedestrians in a scene into small subsets called groups based on their trajectories. Most of previous methods use only a single model for representing a relationship between trajectories of pedestrians who belong to the same group. However, such relationship would vary depending on the activity state (e.g. walking together, approaching, splitting, and so on) of the group. In this paper, we propose a novel group detection method which can cope with a variation of groups' activity state. The proposed method constructs different models for each activity state in order to appropriately evaluate the relationship of pedestrians' trajectories. In addition, our method regards groups' activity state as hidden variables and estimates their probability distributions, which is used for integrating the constructed models. The proposed method outperforms existing methods in the experiment on the public dataset.

［特別招待講演］メディアクローン攻撃を防御するコミュニケーション系－構想・チャレンジ・アプローチ－ 招待

馬場口登

電子情報通信学会技術研究報告   116 ( 497, CQ2016-115 )   25 - 30   2017年3月

　詳細を見る

2017年3月6～7日コミュニケーションクオリティ研究会 (CQ) 於：九州大学大橋キャンパス5号館

A Framework of Privacy-Preserving Image Recognition for Image-Based Information Services 査読

K. Fujii, K. Nakamura, N. Nitta, N. Babaguchi

MULTIMEDIA MODELING (MMM 2017), PT I   10132   40 - 52   2017年

　詳細を見る

Nowadays mobile devices such as smartphones are widely used all over the world. Moreover, the performance of image recognition has dramatically increased by deep learning technologies. From these backgrounds, we think that the following scenario of information services could be realized in the near future: users take a photo and send it to a server, who recognizes the location in the photo and returns the users some information about the recognized location. However, this kind of client-server-based image recognition can cause a privacy issue because image recognition results are sometimes privacy sensitive. To tackle the privacy issue, in this paper, we propose a novel framework for privacy-preserving image recognition in which the server cannot uniquely identify the recognition result but users can do so. An overview of the proposed framework is as follows: First users extract a visual feature from their taken photo and transform it so that the server cannot uniquely identify the recognition result. Then users send the transformed feature to the server, who returns a candidate set of recognition results to the users. Finally, the users compare the candidates and the original visual feature for obtaining the final result. Our experimental results demonstrate the effectiveness of the proposed framework.

Effect of Junk Images on Inter-concept Distance Measurement: Positive or Negative? 査読

Y. Nagasawa, K. Nakamura, N. Nitta, N. Babaguchi

MULTIMEDIA MODELING, MMM 2017, PT II   10133   173 - 184   2017年

　詳細を見る

In this paper, we focus on the problem of inter-concept distance measurement (ICDM), which is a task of computing the distance between two concepts. ICDM is generally achieved by constructing a visual model of each concept and calculating the dissimilarity score between two visual models. The process of visual concept modeling often suffers from the problem of junk images, i.e., the images whose visual content is not related to the given text-tags. Similarly, it is naively expected that junk images also give a negative effect on the performance of ICDM. On the other hand, junk images might be related to its text-tags in a certain (non-visual) sense because the text-tags are given by not automated systems but humans. Hence, the following question arises: Is the effect of junk images on the performance of ICDM positive or negative? In this paper, we aim to answer this non-trivial question from experimental aspects using a unified framework for ICDM and junk image detection. Surprisingly, our experimental result indicates that junk images give a positive effect on the performance of ICDM.

Tag Chat: A Tag-Based Past Topics Recollection Support System. 査読

J. Itou, R. Tanaka, J. Munemori, N. Babaguchi

Proc. of 9th International Conference on Collaboration Technologies (CollabTech 2017)   LNCS 10397   29 - 36   2017年

Long-term people reidentification using anthropometric signature 査読

M. Hasan, N. Babaguchi

2016 IEEE 8th International Conference on Biometrics Theory, Applications and Systems, BTAS 2016   2016年12月

　詳細を見る

Anthropometric biometrics are the most suitable features for long-term people reidentification. However, feature selection is still a major problem in the anthropometric biometrics literature. In this paper, we aim at improving people reidentification by enhancing feature selection. Based on a statistical analysis of body measurements on a large-scale dataset, a new anthropometric signature is introduced. The proposed signature describes both the size and shape of human body at specific anatomical landmarks. While size is measured by Euclidian distance between four skeleton joint pairs, shape is described by the surface distance along four circular body parts. A novel algorithm is proposed to automatically segment the circular body parts from the subject point cloud using body geometry, cylindrical fitting and soft clustering. The overall system is evaluated on two public datasets using CMC and nAUC metrics. Experimental results show the effectiveness of our method compared to state of the art.

Hierarchical detection of rectangles in images 査読

M. Hasan, M. Abdellatif, N. Babaguchi

First International Workshop on Pattern Recognition   7 pages   2016年7月

Privacy Protection for Social Video via Background Estimation and CRF-based Videographer’s Intention Modeling 査読

Y. Nakashima, N. Babaguchi, J. Fan

IEICE Transactions on Information and Systems   E99D ( 4 )   1221 - 1233   2016年4月

Evaluating Protection Capability for Visual Privacy Information 査読

Y. Nakashima, T. Ikeno, N. Babaguchi

IEEE Security and Privacy   14 ( 1 )   55 - 61   2016年1月

　詳細を見る

One way to prevent privacy intrusion is by blurring or blocking out facial images using image processing. However, this technique's effectiveness depends on viewers' familiarity with the subjects as well as on the subjects' conspicuousness.

Real-World Observation Extraction from Microblog based on Word Associative Relations 査読

Masato Yoshitake, Naoko Nitta, Noboru Babaguchi

2016 IEEE SECOND INTERNATIONAL CONFERENCE ON MULTIMEDIA BIG DATA (BIGMM)   450 - 455   2016年

　詳細を見る

A large number of real-world observations by social sensors all over the world can be obtained from various social networking services. Especially, observations covering miscellaneous areas of interest are posted to Twitter as short text messages. Our goal is to extract a wide range of observations related to the target of interest specified by the user from Twitter regardless of their popularity. Assuming that the related observations are likely to contain words people often associate with each other, the associative relations among words are learned from the past messages. When a user gives a keyword representing his/her current interest, recent related observations are extracted based on their word frequency distributions in terms of the associative relations to the keyword.

Facial expression preserving privacy protection using image melding 査読

Yuta Nakashima, Tatsuya Koyama, Naokazu Yokoya, Noboru Babaguchi

Proceedings - IEEE International Conference on Multimedia and Expo   2015-   1 - 6   2015年8月

　詳細を見る

An enormous number of images are currently shared through social networking services such as Facebook. These images usually contain appearance of people and may violate the people's privacy if they are published without permission from each person. To remedy this privacy concern, visual privacy protection, such as blurring, is applied to facial regions of people without permission. However, in addition to image quality degradation, this may spoil the context of the image: If some people are filtered while the others are not, missing facial expression makes comprehension of the image difficult. This paper proposes an image melding-based method that modifies facial regions in a visually unintrusive way with preserving facial expression. Our experimental results demonstrated that the proposed method can retain facial expression while protecting privacy.

Owner authentication for mobile devices using motion gestures based on multi-owner template update 査読

Shigeki Karita, Kumi Nakamura, Kazuhiro Kono, Yoshimichi Ito, Noboru Babaguchi

2015 IEEE International Conference on Multimedia and Expo Workshops, ICMEW 2015   2015年7月

　詳細を見る

This paper proposes a template updating method for improving authentication accuracy in behavioral biometric authentication with hand/arm motion gestures for mobile devices. We introduce an extended version of the standard K-Medoids based clustering algorithm called supervised K-Medoids, which can handle with 2-class data such as positive samples and negative samples. Using the supervised K-Medoids, the template corresponding to each owner is selected as the one that is the most identifiable as the actual owner, and, at the same time, the most distinguishable from the others. Therefore, our method can decrease False-Rejection-Rate (FRR) and False-Acceptance-Rate (FAR) simultaneously, compared to the conventional work that is based on the template update with only the owner's data to decrease FRR. Our template update with multi-owner data attains Equal-Error-Rate (EER) of 5.2% whereas the conventional template update method with owner's own data results in 12.0% when 10 subjects authenticate with gestures for 10 days.

Temporal Spotting of Human Actions from Videos Containing Actor's Unintentional Motions 査読

K. Hara, K. Nakamura, N. Babaguchi

Proc. IEEE International Conference on Multimedia and Expo (ICME2015)   2015年7月

［招待講演］視覚的プライバシー情報の保護と利用－顔を写されて大丈夫？－ 招待

馬場口登

2015年5月

　詳細を見る

中京大学コロキウム 於：中京大学