Adapted Generative Models For Face Verification

It has been shown previously that systems based on local features and relatively complex generative models, namely 1D Hidden Markov Models (HMMs) and pseudo-2D HMMs, are suitable for face recognition (here we mean both identification and verification). Recently a simpler generative model, namely the Gaussian Mixture Model (GMM), was also shown to perform well. In this paper we first propose to increase the performance of the GMM approach (without sacrificing its simplicity) through the use of local features with embedded positional information; we show that the performance obtained is comparable to 1D HMMs. Secondly, we evaluate different training techniques for both GMM and HMM based systems. We show that the traditionally used Maximum Likelihood (ML) training approach has problems estimating robust model parameters when there is only a few training images available; we propose to tackle this problem through the use of Maximum a Posteriori (MAP) training, where the lack of data problem can be effectively circumvented; we show that models estimated with MAP are significantly more robust and are able to generalize to adverse conditions present in the BANCA database.

User Authentication via Adapted Statistical Models of Face Images

Samy Bengio

It has been previously demonstrated that systems based on local features and relatively complex statistical models, namely 1D Hidden Markov Models (HMMs) and pseudo-2D HMMs, are suitable for face recognition. Recently, a simpler statistical model, namely the Gaussian Mixture Model (GMM), was also shown to perform well. In much of the literature devoted to these models, the experiments were performed with controlled images (manual face localization, controlled lighting, background, pose, etc.); however, a practical recognition system has to be robust to more challenging conditions. In this article we evaluate, on the relatively difficult BANCA database, the performance, robustness and complexity of GMM and HMM based approaches, using both manual and automatic face localization. We extend the GMM approach through the use of local features with embedded positional information, increasing performance without sacrificing its low complexity. Furthermore, we show that the traditionally used Maximum Likelihood (ML) training approach has problems estimating robust model parameters when there is only a few training images available; considerably more precise models can be obtained through the use of Maximum a Posteriori (MAP) training. We also show that face recognition techniques which obtain good performance on manually located faces do not necessarily obtain good performance on automatically located faces, indicating that recognition techniques must be designed from the ground up to handle imperfect localization. Finally, we show that while the pseudo 2D HMM approach has the best overall performance, authentication time on current hardware makes it impractical; the best trade-off in terms of authentication time, robustness and discrimination performance is achieved by the extended GMM approach.

IDIAP2004

Bayesian networks as generative models for face recognition

Guillaume Heusch

In this thesis, we address the problem of face modelling by using dedicated statistical generative models, with an application to the face authentication task. Face authentication consists in either accepting or rejecting a user's claim supported by its face image. Classical generative models, such as Gaussian Mixture Models (GMM), Hidden Markov Models (HMM) and their variants have been proved to be successful to tackle this problem. However, these models are not appropriate to the structure of the observed data. In particular, these models implicitly assume independence between features extracted from the image, which is obviously not true in the case of a face. We thus propose new generative models, based on Bayesian Networks, and especially tailored to deal with the object we have to handle: the face. Actually, we would like to exploit as much as a priori knowledge as possible. For this purpose, Bayesian Networks provide an intuitive framework: they allow to encode causal relationships between different kind of random variables, thus enabling to express correlations between different source of information. As a first step, we thus propose a model acting on local observations extracted around salient facial features, which is designed to capture relationships among these pieces of information. The proposed model is shown to be competitive with state-of-the-art approaches based on generative models when applied to the authentication task. At the same level of performance, it is also less complex and thus less time consuming than previous approaches. Besides, and as opposed to classical models, meaningful information could be retrieved from the proposed model. We then extend this model using other sources of information as complementary clues to local features extracted from grayscale face images. Indeed, cognitive studies in face recognition showed that human beings are using various information such as shape, low-resolution representation of the face and also skin color to recognize an individual. We thus proposed new models taking these information into account, and apply them to our authentication problem.

EPFL2010

Bayesian networks as generative models for face recognition

Guillaume Heusch

EPFL2010