1,721,225 research outputs found
Video textures exploiting symmetric movements
No full textIn this paper, an extension to the traditional method for creating video textures is presented. By exploiting the symmetric properties of the frames in a given video sequence, a new endless video stream or video loop is generated by reversing video playback when appropriate. We aim at a larger set of possible input video's for video texture creation, including captured motions containing no smooth transitions (eg. turbulent plant movements). To achieve this, good turn points, instead of transitions are extracted from the video data, after which a simple algorithm synthesizes a new video sequence while optimally exploiting the original frame data. Visual artifacts caused by changed lighting conditions are reduced significantly.We gratefully express our gratitude to the European Fund for Regional Development (ERDF), the Flemish Government and the Flemish Interdisciplinary institute for Broadband Technology (IBBT), which are kindly funding part of the research reported in this paper
Genetic Learning for Designing Sim-to-Real Data Augmentations
No full textData augmentations are useful in closing the sim-to-real domain gap when training on synthetic data. This is because they widen the training data distribution, thus encouraging the model to generalize better to other domains. Many image augmentation techniques exist, parametrized by different settings, such as strength and probability. This leads to a large space of different possible augmentation policies. Some policies work better than others for overcoming the sim-to-real gap for specific datasets, and it is unclear why. This paper presents two different interpretable metrics that can be combined to predict how well a certain augmentation policy will work for a specific sim-to-real setting, focusing on object detection. We validate our metrics by training many models with different augmentation policies and showing a strong correlation with performance on real data. Additionally, we introduce GeneticAugment, a genetic programming method that can leverage these metrics to automatically design an augmentation policy for a specific dataset without needing to train a model
Genetic Learning for Designing Sim-to-Real Data Augmentations Source Code
No full textWe analyze the benefit of data augmentations for overcoming the sim-to-real gap and use the results to develop a genetic learning algorithm for finding augmentation policies
Video textures exploiting symmetric movements
No full textIn this paper, an extension to the traditional method for creating video textures is presented. By exploiting the symmetric properties of the frames in a given video sequence, a new endless video stream or video loop is generated by reversing video playback when appropriate. We aim at a larger set of possible input video's for video texture creation, including captured motions containing no smooth transitions (eg. turbulent plant movements). To achieve this, good turn points, instead of transitions are extracted from the video data, after which a simple algorithm synthesizes a new video sequence while optimally exploiting the original frame data. Visual artifacts caused by changed lighting conditions are reduced significantly.We gratefully express our gratitude to the European Fund for Regional Development (ERDF), the Flemish Government and the Flemish Interdisciplinary institute for Broadband Technology (IBBT), which are kindly funding part of the research reported in this paper
Improving the Runtime Performance of Non-linear Mixed-Effects Model Estimation
No full textNon-linear mixed effects models (NLMEM) are frequently used in drug development for pharmacokinetic (PK) and pharmacokinetic-pharmacodynamic (PK-PD) analyses. Parameter estimation for these models can be time-consuming due to the need for numerical integration. Additionally, the structural model is often expressed using differential equations requiring computationally intensive time-stepping ODE solvers. Overall, this often leads to long computation times in the order of hours or even days. Combining the right mathematical tools as well as techniques from computer science, the computational cost can be significantly reduced. In this paper, several approaches are detailed for improving the performance of parameter estimation for NLMEM. Applying these, often easy, techniques can lead to an order of magnitude speedup
Analysis of Training Object Detection Models with Synthetic Data
No full textRecently, the use of synthetic training data has been on the rise as it offers correctly labelled datasets at a lower cost. The downside of this technique is that the so-called domain gap between the real target images and synthetic training data leads to a decrease in performance. In this paper, we attempt to provide a holistic overview of how to use synthetic data for object detection. We analyse aspects of generating the data as well as techniques used to train the models. We do so by devising a number of experiments, training models on the Dataset of Industrial Metal Objects (DIMO). This dataset contains both real and synthetic images. The synthetic part has different subsets that are either exact synthetic copies of the real data or are copies with certain aspects randomised. This allows us to analyse what types of variation are good for synthetic training data and which aspects should be modelled to closely match the target data. Furthermore, we investigate what types of training techniques are beneficial towards generalisation to real data, and how to use them. Additionally, we analyse how real images can be leveraged when training on synthetic images. All these experiments are validated on real data and benchmarked to models trained on real data. The results offer a number of interesting takeaways that can serve as basic guidelines for using synthetic data for object detection. Code to reproduce results is available at https://github.com/EDM-Research/DIMO_ObjectDetection
Reflections on enhancing interaction in surface computing using real-world objects
No full textThe popularity of interactive surfaces has increased significantly in the past years. The development of several interactive surfaces has stimulated research on interaction techniques, group collaboration, information visualization techniques, technology alternatives, ... Furthermore, a number of applications were realized, mostly focusing on sorting (e.g. pictures), planning (e.g. agenda's, projects, GIS), sketching, brainstorming or games. Here, touch input and gestures form the principal method of interaction, whereas additional input devices are rather seldom used. On the other hand, a number of researchers have been working on tangible user interfaces (TUI) in the past decade. Their research has led to quite some innovative interfaces in a wide range of application areas. Working with tangible I/O objects is experienced as quite intuitive, notwithstanding the fact that many objects are newly created artifacts. These objects have form factors of their real-world counterpart (if any) and thus help facilitating embodied interaction. A research challenge, situated at the crossroads of tangible user interfaces and interactive surfaces, is the interaction with real-world objects on interactive surfaces. Since affordances and behavior of real-world objects are familiar to the target audience of an application (and hence can substantially improve user experience), it should be beneficial to use real world objects in applications of surface computing
Design for accessibility: enabling a motion-impaired artist to control a 3D virtual world
No full textThe application of 3D virtual worlds is not restricted to immersive or desktop
virtual environments. Concepts and techniques that became mature in Virtual
Reality applications are now finding their way in artistic performances. This paper
presents an example of the mutual inspiring effect of art and technology, where the
origin and nature of technology itself is used to come to a new artistic solution and
language. More specific, we describe our approach to give a disabled person the
opportunity to play a major role in artistic performances. In the end, the
performance should not only be appealing for the attendance, but especially for the
actors involved in the show. The wheelchair of the motion-impaired actor is
enhanced by non-intrusive technology, allowing him to navigate in and interact
with a 3D virtual world. This way, he controls the performance by determining the
images perceived by the attendance. The intriguing 3D virtual worlds are presented
to the public via a projection system
Design for accessibility: enabling a motion-impaired artist to control a 3D virtual world
No full textThe application of 3D virtual worlds is not restricted to immersive or desktop
virtual environments. Concepts and techniques that became mature in Virtual
Reality applications are now finding their way in artistic performances. This paper
presents an example of the mutual inspiring effect of art and technology, where the
origin and nature of technology itself is used to come to a new artistic solution and
language. More specific, we describe our approach to give a disabled person the
opportunity to play a major role in artistic performances. In the end, the
performance should not only be appealing for the attendance, but especially for the
actors involved in the show. The wheelchair of the motion-impaired actor is
enhanced by non-intrusive technology, allowing him to navigate in and interact
with a 3D virtual world. This way, he controls the performance by determining the
images perceived by the attendance. The intriguing 3D virtual worlds are presented
to the public via a projection system
EDM-Research/DIMO_ObjectDetection: v1.0
No full textObject detection for the DIMO dataset. Uses the Mask-RCNN model. This is the official implementation of Analysis of Training Object Detection Models with Synthetic Data, published in BMVC: British Machine Vision Conference, 2022.
Source code for the following scientific publication:
Vanherle, B., Moonen, S., Van Reeth, F., and Michiels, N. (2022). Analysis of Training Object Detection Models with Synthetic Data. 33rd British Machine Vision Conference 2022, BMVC 2022, London, UK, November 21-24, 2022. Retrieved from https://bmvc2022.mpi-inf.mpg.de/0833.pdfPILS SBO: Product Inspection with Little Supervision. Flanders Make (Belgium). awardNumber:null. 02ndjfz59BOF Special Research Fund. Hasselt University. awardNumber:null. 10.13039/50110000955
- …
