Publicación:
RGB-D-Fusion: Image Conditioned Depth Diffusion of Humanoid Subjects

Cargando...
Miniatura
Fecha
2023-09-04
Editor/a
Director/a
Tutor/a
Coordinador/a
Prologuista
Revisor/a
Ilustrador/a
Derechos de acceso
info:eu-repo/semantics/openAccess
Título de la revista
ISSN de la revista
Título del volumen
Editor
IEEE Xplore
Proyectos de investigación
Unidades organizativas
Número de la revista
Resumen
We present RGB-D-Fusion, a multi-modal conditional denoising diffusion probabilistic model to generate high resolution depth maps from low-resolution monocular RGB images of humanoid subjects. Accurately representing the human body in 3D is a very active research field given its wide variety of applications. Most 3D reconstruction algorithms rely on depth maps, either coming from low-resolution consumer-level depth sensors, or from monocular depth estimation from standard images. While many modern frameworks use VAEs or GANs for monocular depth estimation, we leverage recent advances in the field of diffusion denoising probabilistic models. We implement a multi-stage conditional diffusion model that first generates a low-resolution depth map conditioned on an image and then upsamples the depth map conditioned on a low-resolution RGB-D image. We further introduce a novel augmentation technique, depth noise augmentation, to increase the robustness of our super-resolution model. Lastly, we show how our method performs on a wide variety of humans with different body types, clothing and poses.
Descripción
Este es el manuscrito aceptado del artículo. La versión registrada fue publicada por primera vez en EEE Access, vol. 11, pp. 99111-99129, 2023, , está disponible en línea en el sitio web del editor: https://doi.org/10.1109/ACCESS.2023.3312017 This is the accepted manuscript of the article. The registered version was first published in EEE Access, vol. 11, pp. 99111-99129, 2023, is available online at the publisher's website: https://doi.org/10.1109/ACCESS.2023.3312017
Categorías UNESCO
Palabras clave
solid modeling, data models, estimation, three-dimensional displays, noise reduction, cameras, probabilistic logic, diffusion processes, deep learning, superresolution, augmented reality, virtual reality, diffusion models, generative deep learning, monocular depth estimation, depth super-resolution, multi-modal, augmented-reality, virtual-reality
Citación
S. Kirch, V. Olyunina, J. Ondřej, R. Pagés, S. Martín and C. Pérez-Molina, "RGB-D-Fusion: Image Conditioned Depth Diffusion of Humanoid Subjects," in IEEE Access, vol. 11, pp. 99111-99129, 2023, doi: https://doi.org/10.1109/ACCESS.2023.3312017
Centro
Facultades y escuelas::E.T.S. de Ingenieros Industriales
Departamento
Ingeniería Eléctrica, Electrónica, Control, Telemática y Química Aplicada a la Ingeniería
Grupo de investigación
Grupo de innovación
Programa de doctorado
Cátedra