3 results
Search Results
Now showing 1 - 3 of 3
Item Comparison of GPU-based Methods for Handling Point Cloud Occlusion(The Eurographics Association, 2021) López, Alfonso; Jurado, Juan Manuel; Padrón, Emilio José; Ogayar, Carlos Javier; Feito, Francisco Ramón; Ortega, Lidia M. and Chica, AntonioThree-dimensional point clouds have conventionally been used along with several sources of information. This fusion can be performed by projecting the point cloud into the image plane and retrieving additional data for each point. Nevertheless, the raw projection omits the occlusion caused by foreground surfaces, thus assigning wrong information to 3D points. For large point clouds, testing the occlusion of each point from every viewpoint is a time-consuming task. Hence, we propose several algorithms implemented in GPU and based on the use of z-buffers. Given the size of nowadays point clouds, we also adapt our methodologies to commodity hardware by splitting the point cloud into several chunks. Finally, we compare their performance through the response time.Item Generation Process of Intrinsic Images Dataset Through Physically-based Rendering(The Eurographics Association, 2021) Rodríguez, Ignacio Moral; López, Alfonso; Jiménez-Perez, J. Roberto; Feito, Francisco R.; Ortega, Lidia; Jurado, Juan M.; Ortega, Lidia M. and Chica, AntonioEl problema denominado Intrinsic Image Decomposition sigue siendo un desafío por resolver en informática gráfica. Aunque el uso de arquitecturas de aprendizaje profundo supondría un avance significativo, los conjuntos de datos de entrenamiento utilizados son aún reducidos. En este estudio se presenta una metodología para la generación de imágenes y su descomposición en varios canales haciendo uso del motor de renderizado Mitsuba2. Para ello, se ha modelado un escenario natural en el que coexisten distintos tipos de vegetación sobre un terreno. En torno a este escenario, se define una trayectoria sobre la que orbita la cámara para generar un conjunto de imágenes desde distintos puntos de vista de forma automática. Como resultado, se proporcionan conjuntos de datos obtenidos a partir de entornos naturales sintéticos formados por las siguientes capas para cada imagen: mapa de normales, iluminación, albedo y mapa de profundidad. Este desarrollo supone un punto de partida para el estudio del cálculo de la iluminación en entornos reales complejos mediante enfoques basados en aprendizaje profundo.Item A GPU-accelerated LiDAR Sensor for Generating Labelled Datasets(The Eurographics Association, 2021) López, Alfonso; Anguita, Carlos Javier Ogayar; Higueruela, Francisco Ramón Feito; Ortega, Lidia M. and Chica, AntonioThis paper presents a GPU-based LiDAR simulator to generate large datasets of ground-truth point clouds. LiDAR technology has significantly increased its impact on academic and industrial environments. However, some of its applications require a large amount of annotated LiDAR data. Furthermore, there exist many types of LiDAR sensors. Therefore, developing a parametric LiDAR model allows simulating a wide range of LiDAR scanning technologies and obtaining a significant number of points clouds at no cost. Beyond their intensity data, these synthetic point clouds can be classified with any level of detail.