Skip to main content

Learning Generative 3D Scene Models for Training and Validating Intelligent Systems

Project description

Training computers to see

Computer vision is an area of artificial intelligence (AI). The goal of computer vision is to equip machines with a visual understanding of their environment, ultimately enabling computers to identify objects in images and videos just like humans do. Much of the recent progress in computer vision builds on machine learning techniques that learn visual representations from large human annotated datasets. However, labeling data for training deep models is expensive and existing photo-realistic simulators do not provide the required variety and fidelity. The EU-funded 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. It will devise algorithms for automatic decomposition of real and synthetic scenes into latent 3D representations capturing geometry, material, light and motion.

Call for proposal

ERC-2019-STG
See other projects for this call

Host institution

EBERHARD KARLS UNIVERSITAET TUEBINGEN
Address
Geschwister-scholl-platz
72074 Tuebingen
Germany
Activity type
Higher or Secondary Education Establishments
EU contribution
€ 1 467 500

Beneficiaries (1)

EBERHARD KARLS UNIVERSITAET TUEBINGEN
Germany
EU contribution
€ 1 467 500
Address
Geschwister-scholl-platz
72074 Tuebingen
Activity type
Higher or Secondary Education Establishments