Estimating 3D hand pose and shape has been gaining a lot of traction in the computer vision community. It has a wide range of applications in the area of robotics, AR/VR and enables technologies such as Microsofts HoloLens, Facebooks Occulus Rift and similar to be interactive. Several topics studying different aspects of this problem are provided. Knowledge from from computer vision and deep learning are leveraged to develop novel methods to tackle these problems. In particular, the focus lies in
1. 3D hand pose estimation (from monocular/stereo RGB or videos)
2. Hand detection from images
3. Dense surface reconstructions (such as hand meshes)
4. Leveraging weakly-/semi-supervised methods to tackle data scarcity to improve the state-of-the-art in hand pose estimation.
Interested students should have a background in computer vision and deep learning. In addition, familiarity with learning frameworks such as PyTorch is required. The projects are research-oriented and if successful can result in a submission to a scientific venue. Therefore students should only apply if they are motivated to work in a research-minded setting. I prefer to work closely with the student, as I believe this brings the most benefits to both of us and the fastest progress in your thesis.
If this fascinating area interests you, I encourage you to contact me via email so we can formulate your thesis topic together. You are also welcome to bring your own ideas to the table.
- Formulate a thesis topic in hand pose estimation based on an outstanding problem in the field.
- Work towards solving the problem.
- If results in scientific novelty, submit to an academic venue.
Adrian Spurr (firstname.lastname@example.org)
hand pose estimation
3D pose estimation
IDEA League Student Grant (IDL)
CLS Student Project (MPG ETH CLS)
ETH Organization's Labels (ETHZ)
Information, Computing and Communication Sciences