Goncalo Pais

Title: Goncalo Pais: Patents in 3D Scene Representation and Neural Rendering

Introduction

Goncalo Pais is an inventor based in Lisbon, Portugal, whose patent work focuses on artificial intelligence methods for representing and rendering three-dimensional scenes. He is associated with Mitsubishi Electric Research Laboratories, Inc. His listed portfolio includes 2 patents.

Latest Patents

Pais’s recent patent work addresses neural rendering and implicit 3D representations.

“Neural radiance field training based on a non-uniform sample of image pixels” describes methods for training a neural network using images of a scene captured from different viewing directions. The approach identifies foreground pixels and samples them more frequently than background pixels, then uses the selected pixels to train the network with radiance values along corresponding viewing rays. This sampling strategy is designed to expedite training.

“Oriented-grid encoder for 3D implicit representation” describes an artificial intelligence system that trains an encoder and decoder to produce an implicit representation of a 3D object. The encoder processes an oriented point cloud, using point locations and surface orientations to create grid-based features at multiple resolutions. The system is designed to capture the object’s shape in a form suitable for 3D scene representation.

Career Highlights

Pais’s listed inventions connect machine learning with computer vision and 3D graphics. His patent work explores ways to make neural scene training more efficient and to encode detailed object geometry from point-cloud data.

Collaborations

Pais is associated with Mitsubishi Electric Research Laboratories, Inc. His listed coworkers include Pedro Miraldo and Marcus Greiff.

Conclusion

Goncalo Pais’s patents highlight approaches to efficient neural radiance field training and multiresolution 3D object representation, reflecting his work at the intersection of artificial intelligence and 3D scene modeling.