Urs Michael Bergmann

Title: Innovations by Urs Michael Bergmann

Introduction

Urs Michael Bergmann is an accomplished inventor based in Berlin, Germany. He has made significant contributions to the field of machine learning, particularly in the area of neural scene representations. His innovative work has garnered attention for its potential applications in various industries.

Latest Patents

Urs Michael Bergmann holds a patent titled "Geometry-free neural scene representations through novel-view synthesis." This patent involves machine learning models that generate geometry-free neural scene representations through efficient object-centric novel-view synthesis. The framework utilizes an encoder model, such as an encoder transformer network, to process one or more RGB images to produce a fully latent scene representation. This representation can then be passed to a decoder model, which synthesizes images in a single forward pass. The use of transformers allows the encoder to learn an attention model that extracts sufficient 3D information from a small set of images, enabling the rendering of novel views with correct projections, parallax, occlusions, and semantics, all without explicit geometry.

Career Highlights

Urs Michael Bergmann is currently employed at Google Inc., where he continues to push the boundaries of innovation in machine learning. His work has been instrumental in advancing the understanding and capabilities of neural scene representations.

Collaborations

Urs has collaborated with notable colleagues, including Mario Lučić and Seyed Mohammad Mehdi Sajjadi. These collaborations have further enriched his research and contributed to the development of cutting-edge technologies.

Conclusion

Urs Michael Bergmann's contributions to the field of machine learning and his innovative patent demonstrate his expertise and commitment to advancing technology. His work is paving the way for new possibilities in neural scene representations and beyond.