The advent of deep learning has led to significant progress in monocular human reconstruction. However, existing representations, such as parametric models, voxel grids, meshes and implicit neural representations, have difficulties achieving high-quality results and real-time speed at the same time. In this paper, we propose Fourier Occupancy Field (FOF), a novel, powerful, efficient and flexible 3D geometry representation, for monocular real-time and accurate human reconstruction. A FOF represents a 3D object with a 2D field orthogonal to the view direction where at each 2D position the occupancy field of the object along the view direction is compactly represented with the first few terms of Fourier series, which retains the topology and neighborhood relation in the 2D domain. A FOF can be stored as a multi-channel image, which is compatible with 2D convolutional neural networks and can bridge the gap between 3D geometries and 2D images. A FOF is very flexible and extensible, \eg, parametric models can be easily integrated into a FOF as a prior to generate more robust results. Meshes and our FOF can be easily inter-converted. Based on FOF, we design the first 30+FPS high-fidelity real-time monocular human reconstruction framework. We demonstrate the potential of FOF on both public datasets and real captured data.

The running time of each component is shown in Table 2.

The advent of deep learning has led to significant progress in monocular human reconstruction.

The advent of deep learning has led to significant progress in monocular human reconstruction. However, existing representations, such as parametric models, voxel grids, meshes and implicit neural representations, have difficulties achieving high-quality results and real-time speed at the same time. In this paper, we propose Fourier Occupancy Field (FOF), a novel, powerful, efficient and flexible 3D geometry representation, for monocular real-time and accurate human reconstruction. A FOF represents a 3D object with a 2D field orthogonal to the view direction where at each 2D position the occupancy field of the object along the view direction is compactly represented with the first few terms of Fourier series, which retains the topology and neighborhood relation in the 2D domain. A FOF can be stored as a multi-channel image, which is compatible with 2D convolutional neural networks and can bridge the gap between 3D geometries and 2D images.

Automated theorem provers take as input the formal statement of a conjecture in a theory described by axioms and lemmas, and try to generate a proof or a counter-example for this conjecture. In the field of geometry, several efficient automated theorem proving approaches have been developed, including algebraic ones such as Wu's method, Gröbner bases method, and semi-synthetic methods such as the area method. In these approaches, typically, the conjecture and the axioms being considered are fixed. However, in mathematical practice, in the context of education and also in mathematical research, the conjecturing and proving activities are not separated but interleaved. The practitioner may try to prove a statement which is valid only assuming some implicit or unknown assumptions, while the list of lemmas and theorem which can be used may not be complete.

The Multiple ANSwer EXtraction system is a framework for interpreting a conjecture with outermost existentially quantified variables as a question, and extracting multiple answers to the question by repetitive calls to a base system that can report the bindings for the variables in one proof of the conjecture. This paper describes the framework and demonstrates its use on an illustrative example.