This paper addresses the lack of a commonly used, standard dataset and established benchmarking problems for physical activity monitoring. A new - recorded from 18 activities performed by 9 subjects, wearing 3 IMUs HR-monitor is created made publicly available. Moreover, 4 classification are benchmarked on dataset, using data processing chain 5 different classifiers. The benchmark shows difficulty tasks exposes challenges

A worldwide movement in advanced manufacturing countries is seeking to reinvigorate (and revolutionize) the industrial and core competencies with use of latest advances information communications technology. Visual computing plays an important role as "glue factor" complete solutions. This article positions visual its intrinsic crucial for Industrie 4.0 provides a general, broad overview points out specific directions scenarios future research.

Physical activity monitoring has recently become an important field in wearable computing research. However, there is a lack of commonly used, standard dataset and established benchmarking problems. In this work, new for physical --- recorded from 9 subjects, wearing 3 inertial measurement units heart rate monitor, performing 18 different activities created made publicly available. Moreover, 4 classification problems are benchmarked on the dataset, using data processing chain 5 classifiers....

Modern large displacement optical flow algorithms usually use an initialization by either sparse descriptor matching techniques or dense approximate nearest neighbor fields. While the latter have advantage of being dense, they major disadvantage very outlier prone as are not designed to find flow, but visually most similar correspondence. In this paper we present a correspondence field approach that is much less and thus better suited for estimation than Our conceptually novel it does...

Establishing correspondences from image to 3D has been a key task of 6DoF object pose estimation for long time. To predict more accurately, deeply learned dense maps replaced sparse templates. Dense methods also improved in the presence occlusion. More recently researchers have shown improvements by learning fragments as segmentation. In this work, we present discrete descriptor, which can represent surface densely. By incorporating hierarchical binary grouping, encode very efficiently....

The paper discusses Archeoguide which offers personalized augmented reality tours of archaeological sites. It uses outdoor tracking, mobile computing, 3D visualization and techniques to enhance information presentation, reconstruct ruined sites, simulate ancient life.

This paper presents the ARCHEOGUIDE project (Augmented Reality-based Cultural Heritage On-site GUIDE). is an IST project, funded by EU, aiming at providing a personalized electronic guide and tour assistant to cultural site visitors. The system provides on-site help Augmented Reality reconstructions of ancient ruins, based on user's position orientation in site, realtime image rendering. It incorporates multimedia database material for on-line access data, virtual visits, restoration...

Learning based approaches have not yet achieved their full potential in optical flow estimation, where performance still trails heuristic approaches. In this paper, we present a CNN patch matching approach for estimation. An important contribution of our is novel thresholded loss Siamese networks. We demonstrate that performs clearly better than existing losses. It also allows to speed up training by factor 2 tests. Furthermore, way calculating features different image scales, which methods....

In scientific evaluation public datasets and benchmarks are indispensable to perform objective assessment. this paper we present a new Comprehensive RGB-D Benchmark for SLAM (CoRBS). contrast state-of-the-art benchmarks, provide the combination of real depth color data together with ground truth trajectory camera 3D model scene. Our novel benchmark allows first time independently evaluate localization as well mapping part systems data. We obtained using an external motion capture system...

Augmented Reality (AR) introduces vast opportunities to the industry in terms of time and therefore cost reduction when utilized various tasks. The biggest obstacle for a comprehensive deployment mobile AR is that current devices still leave much be desired concerning computational graphical performance. To improve this situation paper we introduce an Edge Computing architecture with aim offload demanding algorithms over local network high-end PC considering real-time requirements AR. As...

Articulated hand pose and shape estimation is an important problem for vision-based applications such as augmented reality animation.In contrast to the existing methods which optimize only joint positions, we propose a fully supervised deep network learns jointly estimate full 3D mesh representation from single depth image.To this end, CNN architecture employed parametric representations i.e. pose, bone scales complex parameters. Then, novel layer, embedded inside our framework, produces...

Neural network machine learning approaches are widely used for object classification or detection problems with significant success. A similar problem specific constraints and challenges is state estimation, dealing objects that consist of several removable adjustable parts. system can detect the current such from camera images be great importance Augmented Reality (AR) robotic assembly maintenance applications. In this work, we present a CNN able to regress pose an in multiple states. We...

Depth completion involves recovering a dense depth map from sparse and an RGB image. Recent approaches focus on utilizing color images as guidance to recover at invalid pixels. However, alone are not enough provide the necessary semantic understanding of scene. Consequently, task suffers sudden illumination changes in (e.g., shadows). In this paper, we propose novel three-branch backbone comprising color-guided, semantic-guided, depth-guided branches. Specifically, color-guided branch takes...

Recent works have shown that unstructured text (doc-uments) from online sources can serve as useful auxiliary information for zero-shot image classification. However, these methods require access to a high-quality source like Wikipedia and are limited single of information. Large Language Models (LLM) trained on web-scale show impressive abilities repurpose their learned knowledge multitude tasks. In this work, we provide novel perspective using an LLM supervision classification model. The...

This paper presents a visual SLAM system that uses both points and lines for robust camera localization, simultaneously performs piece-wise planar reconstruction (PPR) of the environment to provide structural map in real-time. One biggest challenges parallel tracking mapping with monocular is keep scale consistent when reconstructing geometric primitives. further introduces difficulties graph optimization bundle adjustment (BA) step. We solve these problems by proposing several run-time...

We present a new visual-inertial tracking device for augmented and virtual reality applications. The paper addresses two fundamental issues of such systems. first one concerns the definition modelling sensor fusion. Much work has been done in this area several models exploiting data gyroscopes linear accelerometers have proposed. However, respective advantages each model particular benefits integration accelerometer filter are still unclear. therefore provides an evaluation different with...

We describe a method for 3D object scanning by aligning depth scans that were taken from around an with Time-of-Flight (ToF) camera. These ToF cameras can measure at video rate. Due to comparably simple technology, they bear potential economical production in big volumes. Our easy-to-use, cost-effective solution, which is based on such sensor, could make technology more accessible everyday users. The algorithmic challenge we face the sensor's level of random noise substantial and there...

3D hand shape and pose estimation from a single depth map is new challenging computer vision problem with many applications. The state-of-the-art methods directly regress meshes 2D images via convolutional neural networks, which leads to artefacts in the estimations due perspective distortions images. In contrast, we propose novel architecture convolutions trained weakly-supervised manner. input our method voxelized map, rely on two representations. first one grid of accurate but does not...

Dense pixel matching is important for many computer vision tasks such as disparity and flow estimation. We present a robust, unified descriptor network that considers large context region with high spatial variance. Our has very receptive field avoids striding layers to maintain resolution. These properties are achieved by creating novel neural layer consists of multiple, parallel, stacked dilated convolutions (SDC). Several these combined form our SDC network. In experiments, we show...