Tutorial at CVPR 2018 Motion Averaging: A Framework for Efficient and Accurate Large-Scale Camera Estimation in 3D Vision.

We are a computer vision group with a primary interest in 3D geometry, using both RGB images (structure-from-motion) and depth cameras.

Our recent work focuses on :

  • robust approaches to motion averaging,
  • large-scale structure-from-motion pipelines,
  • methods for improving and extending the capabilities of depth cameras,
  • using depth cameras for large-scale dense reconstructions.

Other problems of recent interest include principled methods for image processing using patches (e.g. denoising) and efficient higher-order grouping.