João Marcelo X. N. Teixeira

Massively Parallel Nearest Neighbor Queries for Dynamic Point Clouds on the GPU

We introduce a parallel algorithm to solve approximate and exact nearest neighbor queries on the GPU, exploiting its massively parallel processing power. Both data structure construction and nearest neighbor queries are performed on the GPU, avoiding memory copies from system memory to device memory. This algorithm achieves real-time performance, enabling its usage in dynamic scenarios, by minimizing the sorting comparisons needed for a large K value. The underlying data structure for spatial subdivision handles 3D points and is based on grid spatial hashing. Users can specify the grid size interactively. Comparisons were done with other nearest neighbor algorithms implemented on both CPU and GPU. Our approach clearly surpasses CPU implementations regarding processing time, while it presents a competitive solution to GPU ones. Real-time results were obtained with ANN searches (K = 10) for data sets up to 163K points and the potential of our algorithm is demonstrated through a point-based rendering application.

Teleoperation Using Google Glass and AR, Drone for Structural Inspection

This paper proposes the use of a wearable device for visualization and control in association with an UAV applied to the structural inspection of buildings. More specifically, an AR. Drone is controlled through head positions and gestures performed by the operator wearing a Google Glass, and the images captured by the drone are visualized on Glasss screen. We discuss the problems that arise when such a solution is developed, along with the limitations that come from todays available technology and how to overcome them.

kD-Tree Traversal Implementations for Ray Tracing on Massive Multiprocessors: A Comparative Study

Current GPU computational power enables the execution of complex and parallel algorithms, such as Ray Tracing techniques supported by kD-trees for 3D scene rendering in real time. This work describes in detail the study and implementation of five different kD-Tree traversal algorithms using the parallel framework NVIDIA Compute Unified Device Architecture (CUDA), in order to point their pros and cons regarding adaptation capability to the chosen architecture. In addition, a new algorithm is proposed by the authors based on this analysis, aiming performance improvement. A performance analysis of the implemented techniques demonstrates that two of these algorithms, once adequately adapted to CUDA architecture, are capable of reaching speedup gains up to 15x when compared to former CPU implementations and up to 4x in comparison to existing and optimized parallel ones. As a consequence, interactive frame rates are possible for scenes with 1376x768 pixels of resolution and 1 million primitives.

An Open-Source Framework for Air Guitar Games

This paper presents an open-source framework for developing guitar-based games using gesture interaction. The goal of this work was to develop a robust platform capable of providing seamless real time interaction, intuitive playability and coherent sound output. Each part of the proposed architecture is detailed and a case study is performed to exemplify its easiness of use. Some tests are also performed in order to validate the proposed platform. The results showed to be successful: all tested subjects could reach the objective of playing a simple song during a small amount of time and the most important, they were satisfied with the experience.

Nearest Neighbor Searches on the GPU

We introduce a GPU grid-based data structure for massively parallel nearest neighbor searches for dynamic point clouds. The implementation provides real-time performance and it is executed on GPU, both grid construction and nearest neighbors (approximate or exact) searches. This minimizes the memory transfer between device and system memories, improving overall performance. The proposed algorithm may be used across different applications with static and dynamic scenarios. Moreover, our data structure supports three-dimensional point clouds and given its dynamic nature, the user can change the data structure’s parameters at runtime. The same applies to the number of neighbors to be found. Performance comparisons were made against previous works, endorsing the benefits of our solution. Finally, we were able to develop a real-time Point-Based Rendering application for validation of the data structure. Its drawbacks and data distribution’s impact on performance were analysed and some directions for further investigation are given.

Open/Closed Hand Classification Using Kinect Data

This work proposes a study over five different hand gesture classifiers using depth and skeleton data from the Kinect sensor. Evaluations of sensibility, specificity and computational costs are performed for the purpose of choosing which methods are the most adequate. In spite of the low computational complexity of the tested classifiers, the results obtained are of similar quality compared to other complex approaches. All tested classifiers have been gathered in an open-source library for hand classification using the Kinect.

Evaluating Sign Language Recognition Using the Myo Armband

The successful recognition of sign language gestures by computer systems would greatly improve communications between the deaf and the hearers. This work evaluates the usage of electromyogram (EMG) data provided by the Myo armband as features for classification of 20 stationary letter gestures from the Brazilian Sign Language (LIBRAS) alphabet. The classification was performed by binary Support Vector Machines (SVMs), trained with a one-vs-all strategy. The results obtained show that it is possible to identify the gestures, but substantial limitations were found that would need to be tackled by further studies.

Glassist: Using Augmented Reality on Google Glass as an Aid to Classroom Management

Google glass is a hands-free, head-mounted intelligent device that can be worn as a wearable computing eyewear. This work discusses its potential in education by proposing an application, the Glassist, aiming at helping teachers management tasks. The application allows teachers to create individual portfolios for students, manage their information and share it with peers. The tool uses Augmented Reality to recognize students face and display relevant information about them. We also discuss some preliminary results regarding its evaluation. To conclude, we point out some future works.

Increasing kinect application development productivity by an enhanced hardware abstraction

Designing and implementing the interaction behavior for body tracking capable systems requires complex modeling of actions and extensive calibration. Being the most recent and successful device for robust interactive body tracking, Microsofts Kinect has enabled natural interaction by the use of consumer hardware, providing detailed and powerful information to designers and developers, but little tooling. To fulfill this lack of adequate tools for helping developers in the prototyping and implementation of such interfaces, we present Kina, a toolkit that makes the development not fully conditional to the existence of a sensor. By providing playback capabilities together with an online movement database, it reduces the physical effort found while performing testing activities.

FPGA infrastructure for the development of augmented reality applications

In this paper, we describe a platform that aims to help developers on the construction of embedded AR applications. The infrastructure is based on Field Programmable Gate Array and enables the creation of hardware based AR systems, contributing to high speed processing and low power consumption. In addition, the platform uses a componentized design model, where the steps needed to develop an AR application consist in combining the necessary components. A number of image processing algorithms have been implemented and some AR applications have been created using this infrastructure in order to demonstrate the platform capabilities. The performance obtained with the hardware version (at 100 MHz) of the algorithms has also been compared with their software counterparts (at 2 GHz).