One of the main research topics at CVLab is the 3D reconstruction of deformable surfaces from a monocular camera. However, accurate and extensive benchmarking of results on real life object motion have not yet been carried out due to the lack of high quality ground truth data. The goal of this project is to develop an accurate dataset for 3D monocular reconstruction, use it to compare the state-of-the-art methods, and make it publicly available.
Two approaches to build such a benchmark are possible, each having their respective advantages and disadvantages:
An additional task is to build an open source interface to the existing methods developed in the lab.
80% implementation, 20% theory
For further information, send an e-mail regarding the project.
Contact: Aydin Varol (office BC 307, tel 3 64 62).