Amr Ahmed

A framework for automatic semantic video annotation

The rapidly increasing quantity of publicly available videos has driven research into developing automatic tools for indexing, rating, searching and retrieval. Textual semantic representations, such as tagging, labelling and annotation, are often important factors in the process of indexing any video, because of their user-friendly way of representing the semantics appropriate for search and retrieval. Ideally, this annotation should be inspired by the human cognitive way of perceiving and of describing videos. The difference between the low-level visual contents and the corresponding human perception is referred to as the ‘semantic gap’. Tackling this gap is even harder in the case of unconstrained videos, mainly due to the lack of any previous information about the analyzed video on the one hand, and the huge amount of generic knowledge required on the other. This paper introduces a framework for the Automatic Semantic Annotation of unconstrained videos. The proposed framework utilizes two non-domain-specific layers: low-level visual similarity matching, and an annotation analysis that employs commonsense knowledgebases. Commonsense ontology is created by incorporating multiple-structured semantic relationships. Experiments and black-box tests are carried out on standard video databases for action recognition and video information retrieval. White-box tests examine the performance of the individual intermediate layers of the framework, and the evaluation of the results and the statistical analysis show that integrating visual similarity matching with commonsense semantic relationships provides an effective approach to automated video annotation.

Automatic semantic video annotation in wide domain videos based on similarity and commonsense knowledgebases

In this paper, we introduce a novel framework for automatic Semantic Video Annotation. As this framework detects possible events occurring in video clips, it forms the annotating base of a video search engine. To achieve this purpose, the system has to able to operate on uncontrolled wide-domain videos. Thus, all layers have to be based on generic features. The aim is to help bridge the “semantic gap“, which is the difference between the low-level visual features and the humans perception, by finding videos with similar visual events, then analyzing their free text annotation to find the best description for this new video using commonsense knowledgebases. Experiments were performed on wide-domain video clips from the TRECVID 2005 BBC rush standard database. Results from these experiments show promising integrity between those two layers in order to find expressing annotations for the input video. These results were evaluated based on retrieval performance.

Keep, lose, change: Prompts for the re-design of product concepts in a focus group setting

Focus groups have traditionally been used in market and design research to obtain group reactions to product concepts. In this article we outline a simple methodological extension to this format, involving a further stage of concept re-design in smaller subgroups facilitated by a professional designer. The method was developed in the context of working with groups of older people on concepts addressing memory, identity and social communication. It is illustrated with reference to the re-design of two seeded concepts and feedback from participants themselves on the experience of taking part.

VisualNet: Commonsense knowledgebase for video and image indexing and retrieval application

The rapidly increasing amount of video collections, available on the web or via broadcasting, motivated research towards building intelligent tools for searching, rating, indexing and retrieval purposes. Establishing a semantic representation of visual data, mainly in textual form, is one of the important tasks. The time needed for building and maintaining Ontologies and knowledge, especially for wide domain, and the efforts for integrating several approaches emphasize the need for unified generic commonsense knowledgebase for visual applications.

Two-layered Blogger identification model integrating profile and instance-based methods

This paper introduces a two-layered framework that improves the result of authorship identification within larger sample numbers of bloggers as compared with earlier work. Previous studies are mainly divided into two categories: profile-based and instance-based methods. Each of these approaches has its advantages and limitations. The two-layered framework presented here integrates the two previous approaches and presents a new solution to a key problem in authorship identification, namely the drop in accuracy experienced as the number of authors increases. The paper begins by illustrating the regular instance-based core model and the investigated features. It then introduces a new psycholinguistic profile representation of authors, presents similarity grouping extraction over profiles, and applies blogger identification utilizing the two-layered approach. The results confirm the improvement introduced by the proposed two-layered approach against our regular classifier, as well as a selected baseline, for an extended number of users.

Compressed video matching: Frame-to-frame revisited

This paper presents an improved frame-to-frame (F-2-F) compressed video matching technique based on local features extracted from reduced size images, in contrast with previous F-2-F techniques that utilized global features extracted from full size frames. The revised technique addresses both accuracy and computational cost issues of the traditional F-2-F approach. Accuracy is improved through using local features, while computational cost issue is addressed through extracting those local features from reduced size images. For compressed videos, the DC-image sequence, without full decompression, is used. Utilizing such small size images (DC-images) as a base for the proposed work is important, as it pushes the traditional F-2-F from off-line to real-time operational mode. The proposed technique involves addressing an important problem: namely the extraction of enough local features from such a small size images to achieve robust matching. The relevant arguments and supporting evidences for the proposed technique are presented. Experimental results and evaluation, on multiple challenging datasets, show considerable computational time improvements for the proposed technique accompanied by a comparable or higher accuracy than state-of-the-art related techniques.

Compact signature-based compressed video matching using dominant color profiles (DCP)

This paper presents a novel technique for efficient and generic matching of compressed video shots, through compact signatures extracted directly without decompression. The compact signature is based on the Dominant Color Profile (DCP), a sequence of dominant colors extracted and arranged as a sequence of spikes in analogy to the human retinal representation of a scene. The proposed signature represents a given video shot with ~490 integer values, facilitating for real time processing to retrieve a maximum set of matching videos. The technique is able to work directly on MPEG compressed videos, without full decompression, as it utilizes the DC-image as a base for extracting color features. The DC-image has a highly reduced size, while retaining most of visual aspects, and provides high performance compared to the full I-frame. The experiments and results on various standard datasets show the promising performance, both the accuracy and the efficient computation complexity, of the proposed technique.

Biometric template update under facial aging

Being biological tissues in nature, all biometric traits undergo aging. Aging has profound effects on facial biometrics as it causes change in shape and texture. However aging remain an under-studied problem in comparison to facial variations due to pose, illumination and expression changes. A commonly adopted solution in the state-of-the-art is the virtual template synthesis for aging and de-aging transformations involving complex 3D modelling techniques. These methods are also prone to estimation errors in the synthesis. Another viable solution is to continuously adapt the template to the temporal variation (aging) of the query data. Though efficacy of template update procedures has been proven for expression, lightning and pose variations, the use of template update for facial aging has not received much attention so far. This paper investigates the use of template update procedures for temporal variance due to the facial aging process. Experimental evaluations on FGNET and MORPH aging database using commercial VeriLook face recognition engine demonstrate that continuous template updating is an effective and simple way to adapt to variations due to the aging process.

Two-layer classification and distinguished representations of users and documents for grouping and authorship identification

Most studies on authorship identification reported a drop in the identification result when the number of authors exceeds 20–25. In this paper, we introduce a new user representation to address this problem and split classification across two layers. There are at least 3 novelties in this paper. First, the two-layer approach allows applying authorship identification over larger number of authors (tested over 100 authors), and it is extendable. The authors are divided into groups that contain smaller number of authors. Given an anonymous document, the primary layer detects the group to which the document belongs. Then, the secondary layer determines the particular author inside the selected group. In order to extract the groups linking similar authors, clustering is applied over users rather than documents. Hence, the second novelty of this paper is introducing a new user representation that is different from document representation. Without the proposed user representation, the clustering over documents will result in documents of author(s) distributed over several clusters, instead of a single cluster membership for each author. Third, the extracted clusters are descriptive and meaningful of their users as the dimensions have psychological backgrounds. For authorship identification, the documents are labelled with the extracted groups and fed into machine learning to build classification models that predicts the group and author of a given document. The results show that the documents are highly correlated with the extracted corresponding groups, and the proposed model can be accurately trained to determine the group and the author identity.

An investigation into web accessibility standards as a practical study with older and disabled citizens

To coincide with the growth of online services, website standards are progressing to evolve to maintain levels of consistency and usability amongst all web design. Another important aspect is also to include as many users and user groups as possible, especially older and disabled users, given the high potential benefits for those users from the online information, advices, and services. The changes brought about by standardising web design can be both large and small. Ideally the changes will be small enough to go unnoticed by users, but our aim was to see whether even the small changes can make reasonable differences for test users.