Djalma Lucio

In this paper, we introduce the Web Orchestra Studio, a set of applications which enables the development of musical concerts for laptop orchestras. We offer an open-ended platform for collective artistic experimentation which can be utilized by experts or non-initiated students. In order to instance some of the platform features, we present a case study describing our participation in the Math Festival activities with the workshop Music, Mathematics and Computers. Fundamentally, with this work we intend to leverage academic debate concerning the interdisciplinary fields of music, computer science and education.

This paper introduces Probabilistic Editing for Virtual Cinematography. It is part of the VR Kino+Theater platform and provides high level authoring tools for cinematic presentations. The director acts as a DJ controlling in real-time an A/V switcher interface that selects the camera views of a theatrical performance in a virtual reality experience.

We propose a novel gating mechanism applied to Semantic Graph Convolutions for 3D applications, named Semantic Graph Attention. Semantic Graph Convolutions learn to capture semantic information such as local and global node relationships, not explicitly represented in graphs. We improve their performance by proposing an attention block to explore channel-wise inter-dependencies. The proposed method performs the unprojection of the 2D points (image) onto their 3D version. We use it to estimate 3d human pose from 2D images. Both 2D and 3D human poses can be represented as structured graphs, exploring their particularities in this context. The attention layer improves the accuracy of skeleton estimation using 58% fewer parameters than state-of-the-art.

The first terrain modeling techniques were introduced in the late 70’s and in early 80’s. They used procedural methods based on fractals to generate geometry. These models are good enough for some applications, but are not geomorphologically correct. Later, some works introduced methods that create terrains simulating physical process and using real data. In this work we discuss a method to generate terrains based on real samples of digital elevation models. The synthesis process is an adaptation of modern texture synthesis algorithms. Our method goes toward creating non-homogeneous terrains using some controllable features. Keywords-Terrain synthesis, exemplar-based approach, real data

Pose estimation is a challenging task in computer vision that has many applications, as for example: in motion capture, in medical analysis, in human posture monitoring, and in robotics. In other words, it is a main tool to enable machines do understand human patterns in videos or images. Performing this task in real-time while maintaining accuracy and precision is critical for many of these applications. Several papers propose real time approaches considering deep neural networks for pose estimation. However, in most cases they fail when considering run-time performance or do not achieve the precision needed. In this work, we propose a new model for real-time pose estimation considering attention modules for convolutional neural networks (CNNs). We introduce a two-dimensional relative attention mechanism for feature extraction in pose machines leading to improvements in accuracy. We create a single shot architecture where both operations to infer key points and part affinity fields share the information. Also, for each stage, we use tensor decompositions to not only reduce dimensionality, but also to improve performance. This allows us to factorize each convolution and drastically reduce the number of parameters in our network. Our experiments show that, with this factorized approach, it is possible to achieve state-of-art performance in terms of run-time while we have a small reduction in accuracy.

Virtual and Augmented Reality is arguably one of the areas of most rapid development in the game industry nowadays. While some of the technologies have matured and different applications consolidated with market acceptance, the next frontier is still being pursued by researchers and the media. In this paper, we contribute to this quest of effective proposals for the new media with experiments of Situated Participatory VR. They give a glimpse of what will be possible in future applications and guidelines regarding relevant authoring aspects.

Mesa-3D is a device for augmented reality experienced in a support surface, traditionally a table. Mesa-3D is part of the PlanoVision system, which was originally intended for real footage from a static point of view experience. This system expands the experience from PlanoVision by allowing the user freedom of movement. Therefore the authoring and deployment tools for this platform must attend the new requirements, such as head tracking and real-time graphics. Blender 3D was used as a case-study on the design and evaluation of such platform.

With the advent of RGBD cameras, like the Kinect sensor, we gain a powerful tool for image processing. The nature of the captured data (i.e. colored image and depth data) could be combined to realize tasks that before were very difficult. In this paper, we will discuss some camera effects done using RGBD images. These effects will be direct, such as image filtering and stereo image creation, or indirect, such as visualization using head tracking. Keywords-RGBD Image; Camera effects; Filtering; Head Tracking; Stereo Image