In multimodal unsupervised image-to-image translation tasks, the goal is to translate an image from the source domain to many images in the target domain. We present a simple method that produces higher quality images than current state-of-the-art while maintaining the same amount of multimodal diversity. Previous methods follow the unconditional approach of trying to map the latent code directly to a full-size image. This leads to complicated network architectures with several introduced hyperparameters to tune. By treating the latent code as a modifier of the convolutional filters, we produce multimodal output while maintaining the traditional Generative Adversarial Network (GAN) loss and without additional hyperparameters. The only tuning required by our method controls the tradeoff between variability and quality of generated images. Furthermore, we achieve disentanglement between source domain content and target domain style for free as a by-product of our formulation. We perform qualitative and quantitative experiments showing the advantages of our method compared with the state-of-the art on multiple benchmark image-to-image translation datasets.

A significant difficulty in teaching programming lies in the transition from novice to intermediate programmer, characterised by the assimilation and use of schemas of standard programming approaches. A significant factor assisting this transition is practice with tasks which develop this schema use. We describe the Summer of Code, a two-week activity for part-time, distance-learning students which gave them some additional programming practice. We analysed their submissions, forum postings, and results of a terminal survey. We found learners were keen to share and discuss their solutions and persevered with individual problems and the challenge overall. 93% respondents rated the activity 3 or better on a 5-point Likert scale (n=58). However, a quarter of participants, mainly those who described themselves as average or poor programmers, felt less confident in their abilities after the activity, though half of these students liked the activity overall. 54% of all participants said the greatest challenge was developing a general approach to the problems, such as selecting appropriate data structures. This is corroborated by forum comments, where students greatly appreciated “think aloud” presentations by faculty tackling the problems. These results strongly suggest that students would benefit from more open-ended practice, where they have to select and design their own solutions to a range of problems.

Small unmanned aerial vehicles (UAVs) are ideal capturing devices for high-resolution urban 3D reconstructions using multi-view stereo. Nevertheless, practical considerations such as safety usually mean that access to the scan target is often only available for a short amount of time, especially in urban environments. It therefore becomes crucial to perform both view and path planning to minimize flight time while ensuring complete and accurate reconstructions. In this work, we address the challenge of automatic view and path planning for UAV-based aerial imaging with the goal of urban reconstruction from multi-view stereo. To this end, we develop a novel continuous optimization approach using heuristics for multi-view stereo reconstruction quality and apply it to the problem of path planning. Even for large scan areas, our method generates paths in only a few minutes, and is therefore ideally suited for deployment in the field. To evaluate our method, we introduce and describe a detailed benchmark dataset for UAV path planning in urban environments which can also be used to evaluate future research efforts on this topic. Using this dataset and both synthetic and real data, we demonstrate survey-grade urban reconstructions with ground resolutions of 1 cm or better on large areas (30 000 m 2 ).

Correlation filter (CF) based trackers have recently gained a lot of popularity due to their impressive performance on benchmark datasets, while maintaining high frame rates. A significant amount of recent research focuses on the incorporation of stronger features for a richer representation of the tracking target. However, this only helps to discriminate the target from background within a small neighborhood. In this paper, we present a framework that allows the explicit incorporation of global context within CF trackers. We reformulate the original optimization problem and provide a closed form solution for single and multi-dimensional features in the primal and dual domain. Extensive experiments demonstrate that this framework significantly improves the performance of many CF trackers with only a modest impact on frame rate.

In this paper, we propose a persistent, robust and autonomous object tracking system for unmanned aerial vehicles (UAVs) called Persistent Aerial Tracking (PAT). A computer vision and control strategy is applied to a diverse set of moving objects (e.g. humans, animals, cars, boats, etc.) integrating multiple UAVs with a stabilized RGB camera. A novel strategy is employed to successfully track objects over a long period, by `handing over the camera' from one UAV to another. We evaluate several state-of-the-art trackers on the VIVID aerial video dataset and additional sequences that are specifically tailored to low altitude UAV target tracking. Based on the evaluation, we select the leading tracker and improve upon it by optimizing for both speed and performance, integrate the complete system into an off-the-shelf UAV, and obtain promising results showing the robustness of our solution in real-world aerial scenarios.

Code Club is a network of after-school programming clubs for primary (US: elementary) schoolchildren, run by technically-competent volunteers in conjunction with (generally technically-unskilled) teachers. The main motivation of Code Club is to inspire children with a sense of fun and achievement for programming and digital creativity. This paper reports on the first year of Code Club in 1000 UK schools. The results were extremely positive, but some children had difficulty understanding the concepts behind the projects.

In this paper, we describe an approach to the generalisation of tools for teaching and learning the skills associated with modelling with diagrams. We have implemented two revision tools that automatically mark and provide feedback on students' attempts at constructing diagrammatic models of given scenarios in different domains. The similarities between these tools and diagrams in other domains suggest that it might be possible to generalise both the marking algorithm and the drawing editor in such a way that new revision tools could be easily generated for new domains. This paper briefly describes the existing revision tools and our approach to automatic marking of diagrams, and discusses how we are approaching the generalisation of our work for application in other domains.

We have analyzed evolution patterns over two and a half years for a system developed using extreme programming. We find that the system shows a smooth pattern of growth overall, that (McCabe) code complexity is low, and that the relative amount of complexity control work (e.g. refactoring) is higher than in other systems we have studied. To interpret these results, we have drawn on qualitative data including the results of an observational study, records of progress and productivity, and comments on our findings from team members.

In this paper, we present an investigation into the development of a framework for the automatic grading (marking) of student submitted course work. We discuss this framework, its structure and its subsystems. Our framework has been developed in the context of the student submission consisting of two components: a design (using the UML methodology) and source code (using the Java programming language). The focus of our framework is upon the consistency between the student code and design. We discuss its context and development and highlight how we can infer structure from the student submission and use this to inform the assessment process. We describe the approach we are currently undertaking to instantiate this framework and apply it to work submitted for assessment by undergraduate computing students.

In this paper, we describe an approach to the grading (marking) of graph-based diagrams. The work is an application of a framework for the computer interpretation of imprecise diagrams (which students produce in response to assessment questions). We describe the interpretation framework and how it has been applied in the domain of entity-relationship diagrams used in data modeling. In our approach to grading, student diagrams are compared to specimen solutions representing correct solutions to an assessment question. The comparison between a student diagram and a specimen solution diagram results in a matching of the two diagrams from which a grade is computed. The comparison matching is used to provide graphical feedback on the correctness of the student diagram. The approach has been applied to a test corpus of 394 diagrams produced in answer to an examination question with some good results: 91% of all automatically produced grades were within 0.5 of a mark of the human graders. However, when supertype-subtype associations are present, the performance drops to around 87%. The paper concludes with a discussion of some applications of the marker.

Application profiles enable "mixing and matching" metadata elements, in order to meet specific requirements for a particular context. As an example, some communities may want to make certain elements mandatory or restrict the value space of a particular element. However, there is much confusion and only limited experience and expertise in the development and deployment of application profiles. That is why the CEN/ISSS Learning Technologies Workshop has decided to develop guidelines on the use of application profiles for (e-)learning. This paper highlights the main results of this work, and was co-authored by the three experts that drafted the CEN document [4].

This paper illustrates how the notion of pattern can be used in the automatic analysis and synthesis of diagrams, applied particularly to the automatic marking of ER-diagrams. The paper describes how diagram patterns fit into a general framework for diagram interpretation and provides examples of how patterns can be exploited in other fields. Diagram patterns are defined and specified within the area of ER-diagrams. The paper also shows how patterns are being exploited in a revision tool for understanding ER-diagrams.

In this paper we present an approach to the computer understanding of diagrams and show how it can be successfully applied to the automatic marking (grading) of student attempts at drawing entity-relationship (ER) diagrams. The automatic marker has been incorporated into a revision tool to enable students to practice diagramming and obtain feedback on their attempts. Categories and Subject Descriptors