Spatial AI Lab Zurich: Advancing 3D Understanding and Computer Vision

Spatial AI Lab – Zurich: Advancing the Frontiers of Computer Vision and AI

This document provides a comprehensive overview of the research conducted at the Spatial AI Lab in Zurich, a key component of Microsoft Research. The lab focuses on cutting-edge advancements in computer vision, artificial intelligence, and related fields, with a particular emphasis on 3D understanding, scene reconstruction, and the development of novel AI models.

Key Research Areas and Publications

The Spatial AI Lab's research output is diverse, spanning various sub-fields of AI and computer vision. The publications listed highlight significant contributions in areas such as:

• 3D Model Fitting and Optimization:

  • The Phong Surface: Efficient 3D Model Fitting Using Lifted Optimization: This work introduces an efficient method for fitting 3D models using lifted optimization, contributing to more accurate and robust 3D reconstruction.
  • Multi-View Optimization of Local Feature Geometry: Explores optimizing local feature geometry across multiple views, crucial for tasks like Structure from Motion (SfM) and 3D reconstruction.
  • Convolutional Occupancy Networks: Presents a novel approach to representing 3D shapes using convolutional networks, enabling detailed and efficient 3D modeling.
  • Online Invariance Selection for Local Feature Descriptors: Focuses on improving the robustness of feature descriptors by dynamically selecting invariant properties.
  • DIST: Rendering Deep Implicit Signed Distance Function With Differentiable Sphere Tracing: Introduces a new method for rendering implicit signed distance functions, enhancing the realism and efficiency of 3D rendering.
  • Why Having 10,000 Parameters in Your Camera Model Is Better Than Twelve: Discusses the benefits of using more complex camera models for improved accuracy in computer vision tasks.

• Multi-Camera Systems and Calibration:

  • Infrastructure-based Multi-Camera Calibration using Radial Projections: Addresses the challenge of calibrating multiple cameras in a fixed infrastructure, essential for large-scale 3D capture.

• Privacy-Preserving AI:

  • Privacy Preserving Structure-from-Motion: Explores methods to perform SfM while preserving user privacy, a critical consideration in data-driven AI research.

• Image Synthesis and Rendering:

  • Geometry-Aware Satellite-to-Ground Image Synthesis for Urban Areas: Develops techniques for synthesizing realistic ground-level images from satellite imagery, leveraging geometric information.
  • Deep Shutter Unrolling Network: Introduces a deep learning approach to unroll images captured with rolling shutters, correcting for motion artifacts.

Featured Researchers and Collaborations

The lab boasts a strong team of researchers, with notable contributions from:

• Marc Pollefeys: A leading figure in computer vision, his work spans 3D reconstruction, SLAM, and geometric deep learning. He serves as the Partner Director of Science for the lab.
• Viktor Larsson: Contributes significantly to multi-view geometry, camera calibration, and optimization techniques.
• Pablo Speciale: Focuses on privacy-preserving computer vision and 3D reconstruction.
• Zhaopeng Cui: Works on deep learning for computer vision, including image synthesis and rendering.

The lab actively collaborates with other Microsoft Research labs and academic institutions, fostering a vibrant research ecosystem.

Lab Information

• Location: Seestrasse 356, Zurich 8038, Switzerland.
• Contact: Information for following the lab on social media (Facebook, X, LinkedIn, YouTube, Instagram) and subscribing to their RSS feed is provided.

Navigation and Filtering

The page offers robust navigation and filtering options for publications, allowing users to refine results by:

• Research Areas: Computer vision, Artificial intelligence, Graphics and multimedia, Systems and networking, Human-computer interaction, Algorithms.
• People: Researchers associated with the publications.
• Publication Types: Inproceedings (Conference), Article (Journal), Miscellaneous.
• Published Date: Options to filter by all dates, past week, past month, past year, or a custom range.

The current view shows publications 51-60 of 130, sorted by 'Most recent'.

Microsoft Research Ecosystem

The content also provides links to various Microsoft Research initiatives, programs, and resources, including:

• Programs & Events: Academic programs, events, conferences, and the Microsoft Research Forum.
• Connect & Learn: Podcasts, blogs, and forums for engagement.
• About: Information on people, careers, labs, and news.
• All Microsoft: Links to various Microsoft products and services across different categories like Global, Tech & Innovation, Industries, Partners, and Resources.

The Spatial AI Lab in Zurich is at the forefront of AI research, contributing valuable advancements in computer vision and 3D understanding, with a commitment to privacy and innovation.