Descripción del proyecto
Entrenar a los ordenadores para ver
La visión artificial es un campo del conocimiento encuadrado dentro de la inteligencia artificial (IA). El objetivo de la visión artificial es proporcionar a las máquinas una comprensión visual de su entorno, permitiendo así a los ordenadores identificar objetos a partir de imágenes y vídeos de la misma forma en la que lo hacen las personas. La mayoría de los avances recientes en visión artificial se fundamentan en técnicas de aprendizaje automático, que aprenden representaciones visuales a partir de grandes conjuntos de datos humanos anotados. Sin embargo, el etiquetado de datos para el entrenamiento de modelos de aprendizaje profundo es caro y los simuladores fotorrealistas existentes no proporcionan la variedad y la fidelidad necesarias. El proyecto LEGO-3D, financiado con fondos europeos, abordará este problema mediante el desarrollo de modelos probabilísticos capaces de sintetizar escenas tridimensionales junto con proyecciones fotorrealistas bidimensionales a partir de puntos de vista arbitrarios y con un control total sobre los elementos de la escena. Diseñará algoritmos para la descomposición automática de escenas reales y artificiales en representaciones tridimensionales latentes que capturen la geometría, el material, la luz y el movimiento.
Objetivo
Recently, the field of computer vision has witnessed a major transformation away from expert designed shallow models towards more generic deep representation learning. However, collecting labeled data for training deep models is costly and existing simulators with artist-designed scenes do not provide the required variety and fidelity. Project LEGO-3D will tackle this problem by developing probabilistic models capable of synthesizing 3D scenes jointly with photo-realistic 2D projections from arbitrary viewpoints and with full control over the scene elements. Our key insight is that data augmentation, while hard in 2D, becomes considerably easier in 3D as physical properties such as viewpoint invariances and occlusion relationships are captured by construction. Thus, our goal is to learn the entire 3D-to-2D simulation pipeline. In particular, we will focus on the following problems:
(A) We will devise algorithms for automatic decomposition of real and synthetic scenes into latent 3D primitive representations capturing geometry, material, light and motion.
(B) We will develop novel probabilistic generative models which are able to synthesize large-scale 3D environments based on the primitives extracted in project (A). In particular, we will develop unconditional, conditioned and spatio-temporal scene generation networks.
(C) We will combine differentiable and neural rendering techniques with deep learning based image synthesis, yielding high-fidelity 2D renderings of the 3D representations generated in project (B) while capturing ambiguities and uncertainties.
Project LEGO-3D will significantly impact a large number of application areas. Examples include vision systems which require access to large amounts of annotated data, safety-critical applications such as autonomous cars that rely on efficient ways for training and validation, as well as the entertainment industry which seeks to automate the creation and manipulation of 3D content.
Ámbito científico
- engineering and technologymechanical engineeringvehicle engineeringautomotive engineeringautonomous vehicles
- natural sciencescomputer and information sciencesartificial intelligencecomputer vision
- natural sciencescomputer and information sciencesartificial intelligencemachine learningdeep learning
- natural sciencesmathematicspure mathematicsgeometry
Palabras clave
Programa(s)
Régimen de financiación
ERC-STG - Starting GrantInstitución de acogida
72074 Tuebingen
Alemania