University of Cambridge > Talks.cam > Rainbow Group Seminars > Where Ideas Take Shapes

Where Ideas Take Shapes

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Rafal Mantiuk.

In this talk, I will provide a summary of high-level ideas behind the most significant research we made over the last decade in the area of Computer Graphics, Computer Vision, Digital Fabrication and Deep Learning.

First, a video-based graphics pipeline to capture surfaces in motion is presented. In the context of full-body human performance digitalization, dynamic surfaces are governed by complex non-rigid behaviors hard to parametrize. The key idea is to rely on dimension reduction and variational principles. In practice, a low-dimensional subspace coupled with generalized barycentric coordinates is used to manipulate space-time deformable geometry. This space-based deformer helps to register template-based deformation driven by multi-view silhouettes.

Next, a hollowing optimization algorithm based on the concept of honeycomb-cells structure for additive manufacturing will be showcased. Nowadays, 3D printing is an ubiquitous technology enabling low-cost and fast prototyping for digital fabrication of real-world objects. The key idea is to infer a stress-based Voronoi tessellation suitable for carving harmonic porous in the solid.

Finally, a few-shot Generative Adversarial Network architecture is discussed to perform deep learning tasks with less amount of training data. The key idea is that the generation improves the classification while the classification improves the generation. Consequently, the model can learn adaptively by switching between full supervised learning and semi-supervised learning. This modified GAN is less restricted in term of input and output.

In closing, the importance of optimality at the heart of computational models for visual computing will be discussed. Also, some potential directions to enrich our field and to improve existing techniques for Capturing, Learning and Printing will be sketched.

This talk is part of the Rainbow Group Seminars series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.

 

© 2006-2019 Talks.cam, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity