Disentangling Factors of Variation for Facial Expression Recognition

By Salah Rifai, Yoshua Bengio, Aaron C. Courville, Pascal Vincent, Mehdi Mirza. 2012

211 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Experimental and Cognitive Psychology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Experimental and Cognitive Psychology

Citations in the corpus

Revisiting Natural Gradient for Deep Networks (2013, methodology): “ore generalization-friendly metric G. Figure 2 describes the results on the Toronto Face Dataset (TFD), where using unlabeled data results in 83.04% accuracy vs 81.13% without. State of the art is 85%Rifai et al. (2012), though this result is obtained by a larger model that is pre-trained. Hyper-par”
Context Encoders: Feature Learning by Inpainting (2016, background): “[36] demonstrate that CNNs can learn to generate novel images of particular object categories (chairs and faces, respectively), but rely on large labeled datasets with examples of these categories.”
Learning deep representations by mutual information estimation and maximization (2018, background): “In generative models that rely on reconstruction (e.g., denoising, variational, and adversarial autoencoders, Vincent et al., 2008; Rifai et al., 2012; Kingma & Welling, 2013; Makhzani et al., 2015), the negative reconstruction error can be related to the MI in the encoder as, where X and Y are…”
Action-Conditional Video Prediction using Deep Networks in Atari Games (2015, neutral): “This result implies that our model learns to distinguish between controlled objects and uncontrolled objects and transform them using disentangled representations (see [25, 24, 37] for related work on disentangling factors of variation).”
Training Deep Neural Networks on Noisy Labels with Bootstrapping (2014, neutral): “, 2014) (Rifai et al., 2012)”
Discovering Hidden Factors of Variation in Deep Networks (2014, background): “[17] proposed a similar penalty over terms in the product between the Jacobians of observed and latent variables with respect to the input.”
Deep Learning of Representations: Looking Forward (2013, methodology): “This can be used to endow regularized auto-encoders with a probabilistic interpretation and to sample from the implicitly learned density models (Rifai et al., 2012b; Bengio et al., 2012; Alain and Bengio, 2013) through some variant of Langevin or Metropolis-Hastings Monte-Carlo Markov chains…”
Learned-Norm Pooling for Deep Feedforward and Recurrent Neural Networks (2013, methodology): “s paper were done using Pylearn2 library (Goodfellow et al., 2013b). Further details about experiments are provided in the Supplementary Material. 1Reported by Goodfellow et al. (2013a). 2Reported by Rifai et al. (2012). 3Reported by Gulcehre and Bengio (2013). 4Reported by Rifai et al. (2011). 5Rep”
A Survey of Inductive Biases for Factorial Representation-Learning (2016, methodology): “A contractive regularization penalty can be used to encourage factors to be insensitive to local directions of variation in the observation space [42].”
Image-to-image translation for cross-domain disentanglement (2018, background): “One of the earliest architectures for learning disentangled representations using deep learning was applied to the task of emotion recognition [38].”
Learning and Selecting Features Jointly with Point-wise Gated Boltzmann Machines (2013, unknown): “Rifai et al. (2012) proposed the contractive discriminative analysis (CDA).”
Learning to Disentangle Factors of Variation with Manifold Interaction (2014, background): “Contractive discriminative analysis (Rifai et al., 2012) also learns groups of task-relevant and irrelevant hidden units using a contractive penalty, but only uses additive interactions between the input and each group of hidden units.”
Deep Region and Multi-label Learning for Facial Action Unit Detection (2016, unknown): “, 48×48 in [22]), we set the input image to 170×170.”
Survey on RGB, 3D, Thermal, and Multimodal Approaches for Facial Expression Recognition: History, Trends, and Affect-related Applications (2016, methodology): “CNNs are used in [144], [145], [146], [147], [148].”
Deep Visual Analogy-Making (2015, methodology): “Disentangling was also considered in the discriminative case in the Contractive Discriminative Analysis model [24].”
FaceNet2ExpNet: Regularizing a Deep Face Recognition Net for Expression Recognition (2016, methodology): “ of 3.1%. The confusion matrix is shown in Fig. 7. The proposed method performs TABLE V THE AVERAGE ACCURACY ON TFD DATASET Method Average Accuracy Gabor + PCA [40] 80.2% Deep mPoT [41] 82.4% CDA+CCA [42] 85.0% disRBM [43] 85.4% bootstrap-recon [44] 86.8% Train From Scratch (BN) 82.5% VGG Fine-Tune ”
Learning Salient Features for Speech Emotion Recognition Using Convolutional Neural Networks (2014, background): “In many situations where labeled data is limited or not available, deep learning is shown to have the capability to generate good features, for example, for facial expression recognition [31] and ASR [43].”
Multimodal Representation Learning via Maximization of Local Mutual Information (2021, unknown): “The definition of usefulness is casedriven [3,5,26].”
Balanced Datasets Are Not Enough: Estimating and Mitigating Gender Bias in Deep Image Representations (2018, background): “Building models under fairness objectives is also related to feature disentangling methods [28, 22, 17, 18, 19].”