Modeling Stylized Character Expressions via Deep Learning

Aneja, Deepali; Colburn, Alex; Faigin, Gary; Shapiro, Linda G.; Mones, Barbara

doi:10.1007/978-3-319-54184-6_9

Cited by 83 publications

(75 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(c) FERG-DB [67], contains 55767 face images from 6 stylized characters with annotated facial expressions. The images for each character are grouped into 7 types of expressions, i.e., anger, disgust, fear, joy, neutral, sadness and surprise.…”

Section: Methodsmentioning

confidence: 99%

Understanding autoencoders with information theoretic concepts

Prı́ncipe

2019

Neural Networks

116

View full text Add to dashboard Cite

Despite their great success in practical applications, there is still a lack of theoretical and systematic methods to analyze deep neural networks. In this paper, we illustrate an advanced information theoretic methodology to understand the dynamics of learning and the design of autoencoders, a special type of deep learning architectures that resembles a communication channel. By generalizing the information plane to any cost function, and inspecting the roles and dynamics of different layers using layer-wise information quantities, we emphasize the role that mutual information plays in quantifying learning from data. We further suggest and also experimentally validate, for mean square error training, three fundamental properties regarding the layer-wise flow of information and intrinsic dimensionality of the bottleneck layer, using respectively the data processing inequality and the identification of a bifurcation point in the information plane that is controlled by the given data. Our observations have direct impact on the optimal design of autoencoders, the design of alternative feedforward training methods, and even in the problem of generalization.

show abstract

Section: Methodsmentioning

confidence: 99%

Understanding autoencoders with information theoretic concepts

Prı́ncipe

2019

Neural Networks

116

View full text Add to dashboard Cite

show abstract

“…An animator created the key poses for each expression, and they were labeled via MT to populate the database initially. The details of the database collection and data pre-processing are given in [1].…”

Section: Methodsmentioning

confidence: 99%

Learning to Generate 3D Stylized Character Expressions from Humans

Aneja¹,

Chaudhuri²,

Colburn

et al. 2018

2018 IEEE Winter Conference on Applications of Computer Vision (WACV)

Self Cite

View full text Add to dashboard Cite

“…e(z i ) are "one-hot" vectors (z i component is 1 and the others 0) of the true labels of sample i = [1 : n].p Z is the empirical distribution of the private labels. Each branch, i.e., qŶ |V and qẐ |X , is trained to minimize the associated cross-entropy loss (respectively (13) and (14)), whereas the autoencoder qX |X is trained to minimize eq. 12(minimizing the cross-entropy loss with respect toŶ predictor while maximizing the new adversarial loss defined with respect to the Z predictor).…”

Section: Presentation Of the Problemmentioning

confidence: 99%

Learning Semi-Supervised Anonymized Representations by Mutual Information

Feutry

Piantanida

Duhamel

2020

ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

This paper addresses the problem of removing from a set of data (here images) a given private information, while still allowing other utilities on the processed data. This is obtained by training concurrently a GAN-like discriminator and an autoencoder. The optimization of the resulting structure involves a novel surrogate of the misclassification probability of the information to remove. Several examples are given, demonstrating that a good level of privacy can be obtained on images at the cost of the introduction of very small artifacts.

show abstract

Modeling Stylized Character Expressions via Deep Learning

Cited by 83 publications

References 37 publications

Understanding autoencoders with information theoretic concepts

Understanding autoencoders with information theoretic concepts

Learning to Generate 3D Stylized Character Expressions from Humans

Learning Semi-Supervised Anonymized Representations by Mutual Information

Contact Info

Product

Resources

About