Large-scale 3D Reconstruction
The goal of this line of research is to develop scalable and robust 3D reconstruction systems. In particular, the laboratory is interested in reconstructing scenes from large photo collections. Given the large size of these input photo collections, the laboratory is interested in developing reconstruction algorithms that can scale with the size of the collection. Also, given that natural scenes pose complex problems that can yield reconstruction systems to fail, the laboratory is interested in developing robust methods that enable systems to overcome these posed difficulties when dealing with images from natural scenes.The laboratory currently works in the following areas:
- Robust estimation: The research team develops estimators that enable systems to recover the geometry of the scene quickly and reliably.
- Structure from motion (SfM): The research team creates efficient novel algorithms that enable SfM systems to scale and produce accurate reconstructions.
Related Publications:
1. C. Sweeney, V. Fragoso, T. Hollerer, M. Turk. Large Scale SfM with the Distributed
Camera Model. IEEE 3DV. 2016.
PDF
2. C. Sweeney, V. Fragoso, T. Hollerer, M. Turk. gDLS: A Scalable Solution to the
Generalized Pose and Scale Problem. IEEE ECCV. 2014.
PDF
3. V. Fragoso, G. Srivastava, A. Nagar, Z. Li, K. Park, M. Turk. Cascade of
Box (CABOX) Filters for Optimal Scale Space Estimation. IEEE CVPR Workshops. 2014.
PDF
4. V. Fragoso, P. Sen, S. Rodriguez, and M. Turk. EVSAC: Accelerating Hypothesis
Generation by Modeling Matching Scores with Extreme Value Theory. IEEE ICCV. 2013.
PDF
5. Victor Fragoso, M. Turk. SWIGS: A Swift Guided Sampling Method. IEEE CVPR. 2013.
PDF
6. V. Fragoso, M. Turk, and J. Hespanha. Locating Binary Features for Keypoint
Recognition using Noncooperative Games. IEEE ICIP 2012.
PDF