Maia Rohm

TU Wien | TU Wien · Interactive Media Systems Group

A novel approach for outlier detection is proposed, called local projections, which is based on concepts of the Local Outlier Factor (LOF) (Breunig et al. in Lof: identifying density-based local outliers. In: ACM sigmod record, ACM, volume 29, pp. 93–104, 2000) and ROBPCA (Hubert et al. in Technometrics 47(1):64–79, 2005). By using aspects of both...

Image retrieval has been an active research domain for over 30 years and historically it has focused primarily on precision as an evaluation criterion. Similar to text retrieval, where the number of indexed documents became large and many relevant documents exist, it is of high importance to highlight diversity in the search results to provide bett...

In real-world application scenarios, the identification of groups poses a significant challenge due to possibly occurring outliers and existing noise variables. Therefore, there is a need for a clustering method which is capable of revealing the group structure in data containing both outliers and noise variables without any pre-knowledge. In this...

In this paper, we present a new dataset that facilitates the comparison of approaches aiming at the diversification of image search results. The dataset was explicitly designed for general-purpose, multi-topic queries and provides multiple ground truth annotations to allow for the exploration of the subjectivity aspect in the general task of divers...

Media content in large repositories usually exhibits multiple groups of strongly varying sizes. Media of potential interest often form notably smaller groups. Such media groups differ so much from the remaining data that it may be worthy to look at them in more detail. In contrast, media with popular content appear in larger groups. Identifying gro...

Media content in large repositories usually exhibits multiple groups of strongly varying sizes. Media of potential interest often form notably smaller groups. Such media groups differ so much from the remaining data that it may be worthy to look at them in more detail. In contrast, media with popular content appear in larger groups. Identifying gro...

In real-world application scenarios, the identification of groups poses a significant challenge due to possibly occurring outliers and existing noise variables. Therefore, there is a need for a clustering method which is capable of revealing the group structure in data containing both outliers and noise variables without any pre-knowledge. In this...

A novel approach for supervised classification analysis for high dimensional and flat data (more variables than observations) is proposed. We use the information of class-membership of observations to determine groups of observations locally describing the group structure. By projecting the data on the subspace spanned by those groups, local projec...

The selection of an appropriate feature set is crucial for the efficient analysis of any media collection. In general, feature selection strongly depends on the data and commonly requires expert knowledge and previous experiments in related application scenarios. Current unsupervised feature selection methods usually ignore existing relationships a...

In this paper, we propose a novel approach for outlier detection, called local projections, which is based on concepts of Local Outlier Factor (LOF) (Breunig et al., 2000) and RobPCA (Hubert et al., 2005). By using aspects of both methods, our algorithm is robust towards noise variables and is capable of performing outlier detection in multi-group...

A powerful data transformation method named guided projections is proposed creating new possibilities to reveal the group structure of high-dimensional data in the presence of noise variables. Utilising projections onto a space spanned by a selection of a small number of observations allows measuring the similarity of other observations to the sele...

A powerful data transformation method named guided projections is proposed creating new possibilities to reveal the group structure of high-dimensional data in the presence of noise variables. Utilising projections onto a space spanned by a selection of a small number of observations allows measuring the similarity of other observations to the sele...

In this paper, we explore the application of an adaptive clustering approach for the diversification of image retrieval results in the context of the MediaEval 2016 Retrieving Diverse Social Images Task. The proposed approach exploits available textual descriptions, the visual content of the images , and a set of common clustering techniques to sel...

This paper provides an overview of the Retrieving Diverse Social Images task that is organized as part of the MediaEval 2016 Benchmarking Initiative for Multimedia Evaluation. The task addresses the problem of result diversification in the context of social photo retrieval where images, meta-data, text information, user tagging profiles and content...

Outliers often reveal crucial information about the underlying data such as the presence of unusual observations that require for in-depth analysis. The detection of outliers is especially challenging in real-world application scenarios dealing with high-dimensional and flat data bearing different subpop-ulations of potentially varying data distrib...

Recurring visual elements in videos commonly represent central content entities, such as main characters and dominant objects. The automated detection of such elements is crucial for various application fields ranging from compact video content summarization to the retrieval of videos sharing common visual entities. Recent approaches for content-ba...

In this work we present an approach for video content representation based on the detection of recurring visual elements or regions. We hypothesize that such elements play a potentially central role in the underlying video sequence. The approach makes use of fundamental intrinsic properties of a video and, thus, it does not make any assumptions abo...

The performance of video genre classification approaches strongly depends on the selected feature set. Feature selection requires for expert knowledge and is commonly driven by the underlying data, investigated video genres, and previous experience in related application scenarios. An alteration of the genres of interest results in reconsideration...

In this paper, we describe our approach for the MediaEval 2015 Retrieving Diverse Social Images Task. The proposed approach exploits available user-generated textual descriptions and the visual content of the images in a combination with common, unsupervised clustering techniques in order to increase the diversification of retrieval results. Prelim...

A large part of media shared on online platforms such as Flickr and YouTube is captured at various social events (e.g. music festivals, exhibitions, and sport events). While it is quite easy to share personal impressions online, it is much more challenging to identify content that is related to the same social event across different platforms. In t...

Outliers often reveal crucial information about the underlying data such as the presence of unusual observations that require for in-depth analysis. The detection of outliers is especially challenging in real-world application scenarios dealing with high-dimensional and flat data bearing different subpopulations of potentially varying data distribu...

A significant part of publicly available photos on the Internet depicts a variety of different social events. In order to organize this steadily growing media content and to make it easily accessible, novel indexing methods are required. Essential research questions in this context concern the efficient detection (clustering), classification, and r...

The immense amount of available video data poses novel requirements for video representation approaches by means of focusing on central and relevant aspects of the underlying story and facilitating the efficient overview and assessment of the content. In general, the assessment of content relevance and significance is a high-level task that usually...

A large part of media shared on online platforms such as Flickr and YouTube is captured at various social events (e.g. music festivals, exhibitions, and sport events]. While it is quite easy to share personal impressions online, it is much more challenging to identify content that is related to the same social event across different platforms. In t...

Feature selection is applied to identify relevant and complementary features from a given high-dimensional feature set. In general, existing filter-based approaches operate on single (scalar) feature components and ignore the relationships among components of multidimensional features. As a result, generated feature subsets lack in interpretability...

This paper describes our contribution to the MediaEval 2014 task on the Synchronization of multi-user Event Media (SEM). We propose two multimodal approaches that employ both visual and time information for the synchronization of different images galleries and for the detections of sub-events. The methods prove robustness in the determination of ti...

In this paper we explore the performance of a generic, unified framework for the retrieval of relevant and diverse images from social photo collections. The approach allows for the easy evaluation of different visual and textual image descriptions, clustering algorithms, and similarity metrics. Preliminary results show strong dependance between the...

This paper describes our contributions to the Social Event Detection (SED) task as part of the MediaEval Benchmark 2014. We first present an unsupervised approach for the clustering of social events that builds solely on provided metadata. Results show that already the use of available time and location information achieves high clustering precisio...

The detection of a specific social event requires for high semantic understanding in the interpretation of particular event characteristics such as its type and location. In many cases, photos capturing different events at the same (or highly similar) locations can hardly be distinguished by each other. Available metadata can provide assistance whe...

Recurring elements in movies contribute significantly to the development of narration, themes, or even mood. The detection of such elements is impeded by the large variance of their visual appearance and usually relies on the experience and attentiveness of the viewer. In this paper, we present a new approach for the automated detection of recurrin...

In this paper, the authors present an approach for video comparison, in which an instantiated framework allows for the easy comparison of different methods that are required at each step of the comparison process. The authors’ approach is evaluated based on a real world scenario of challenging video data of archive documentaries. In this paper, the...

This article presents methods for the automatic retrieval of motion and motion compositions in movies. We introduce a user-friendly sketch-based query interface that enables the user to describe desired motion compositions. Based on this abstract description, a tolerant matching scheme extracts shots from a movie with a similar composition. We inve...

We present a vision-based approach to ancient coins’ identification. The approach is a two-stage procedure. In the first stage an invariant shape description of the coin edge is computed and matching based on shape is performed. The second stage uses preselection by the first stage in order to refine the matching using local descriptors. Results fo...

This article outlines issues related to automated film analysis for archived documentaries and explores examples of the applicability of existing content-based retrieval methods.

Scene segmentation is a crucial task in the structural analysis of film. State-of-the-art scene segmentation algorithms usually target fiction films (e.g. Hollywood films). Documentaries (especially artistic archive documentaries) follow different montage rules than fiction films and consequently require specialized approaches for scene segmentatio...

In this paper, the authors present an approach for video comparison, in which an instantiated framework allows for the easy comparison of different methods that are required at each step of the comparison process. The authors' approach is evaluated based on a real world scenario of challenging video data of archive documentaries. In this paper, the...

The contribution of this paper consists of a framework for video comparison that allows for the analysis of different movie versions. Furthermore, a second contribution is an evaluation of state-of-the-art, local feature-based approaches for content-based video retrieval in a real world scenario. Eventually, the experimental results show the outsta...

We propose a novel clustering scheme for spatio-temporal segmentation of sparse motion fields obtained from feature tracking. The approach allows for the segmentation of meaningful motion components in a scene, such as short- and long-term motion of single objects, groups of objects and camera motion. The method has been developed within a project...

In this paper we focus on a novel issue in the field of video retrieval stemming from film analysis, namely the investigation of film montage patterns. For this purpose it is first necessary to reconstruct the original film sequences, i.e. the camera takes. For the decision whether or not two shots occurring anywhere in a film stem from the same ta...

Reliable object identification is an essential task in the process of recognizing and tracing stolen cultural heritage. We investigate the feasibility of using computer-aided identification of ancient coins to search for a given coin on the Internet or in a digital repository. Because a coin's shape is a unique feature, we first apply a shape descr...

Reliable object identification is an essential task in the process of recognition and traceability of stolen cultural heritage. Existing approaches for object recognition focus mainly on object classification. However, they are not sufficient to identify a given object among hundreds of objects of the same class. In this paper, we investigate the f...

Numismatics deals with various historical aspects of the phenomenon money. Fundamental part of a numismatists work is the identification and classification of coins according to standard reference books. The recognition of ancient coins is a highly complex task that requires years of experience in the entire field of numismatics. To date, no optica...

This contribution investigates the content-based feature extraction methods used in visual information retrieval, focusing on concepts that are employed for the semantic representation of media content. The background part describes the building blocks of feature extraction functions. Since numerous methods have been proposed we concentrate on the...

The experiments described in this paper indicate that under certain conditions content-based features are not required for efficient user-centred image retrieval in small media collections. The importance of feature selection drops dramatically if classification is used for retrieval (e.g. if Support Vector Machines are used) and only little user f...

Illegal trade and theft of coins appears to be a major part of the illegal antiques market. Image based recognition of coins could substantially contribute to fight against it. Central component in the permanent identification and traceability of coins is the underlying classification and identification technology. However, currently available algo...

Illegal trade and theft of coins appears to be a major part of the illegal antiques market. Image based recognition of coins could substantially contribute to fight against it. Central component in the permanent identification and traceability of coins is the underlying classification and identification technology. However, currently available algo...

Illegal trade and theft of coins appears to be a major part of the illegal antiques market. Image based recognition of coins could substantially contribute to fight against it. Central component in the permanent identification and traceability of coins is the underlying classification and identification technology. The first step of a computer aide...

Mobile devices are becoming more and more important in the context of e-learning. This requires appropriate models for structuring and delivering content to be used on various devices. Different technical characteristics of devices as well as different needs of learners require specific approaches. In this paper we propose a model for structuring c...

This paper gives an overview on the recently granted EU project COINS which stands for COmbatting Illicit Numismatic Sales. The proposed project aims to substantially contribute to the fight against illegal trade and theft of coins which appears to be a major part of the illegal antiques market. Central component in the permanent identification and...