Our research focuses on learning the low-dimensional embeddings of face images.
Subspace learning for computer vision applications has recently generated a significant amount of scientific research. This research has been primarily motivated by the development of a multitude of techniques for the efficient analysis of high-dimensional data via non-linear dimensionality reduction.
These techniques have provided valuable tools for understanding and capturing the intrinsic non-linear structure of visual data encountered in many important machine vision problems. At the same time, there has been a substantially increasing interest in related applications such as appearance-based object/face recognition.
The group's research concentrates on robust kernel-based extensions to linear subspace analysis as well as manifold learning techniques.
Visualization of FreyFaces using Locally Linear Embedding (left figure) and our learning algorithm (right figure)
Maja Pantic, Baris Gecer, Athanasios Papaioannou, Georgios Tzimiropoulos, Yujiang Wang
G. Tzimiropoulos, S. Zafeiriou, M. Pantic. IEEE Transactions on Pattern Analysis and Machine Intelligence. 34(12): pp. 2454 - 2466, 2012.
S. Zafeiriou, G. Tzimiropoulos, M. Petrou, T. Stathaki. IEEE Transactions on Neural Networks and Learning Systems (accepted for publication). 2011.
G. Tzimiropoulos, S. Zafeiriou, M. Pantic. Proceedings of IEEE International Conference on Automatic Face and Gesture Recognition (FG'11). Santa Barbara, CA, USA, pp. 553 - 558, March 2011.