Separating Style and Content

By Josh Tenenbaum, William T. Freeman. 1996

How researchers see this paper

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

Citations in the corpus

Interpretable Basis Decomposition for Visual Explanation (2018, methodology): “Other decomposition methods such as Bilinear models [23] and Isomap [22] are also used to discover meaningful subspaces and structure in the data.”
Beyond Bilinear: Generalized Multimodal Factorized High-Order Pooling for Visual Question Answering (2017, methodology): “In contrast to linear pooling, bilinear pooling [11] has recently been used to integrate different CNN features for fine-grained image recognition [12].”
Multi-modal Factorized Bilinear Pooling with Co-attention Learning for Visual Question Answering (2017, methodology): “In contrast to linear pooling, bilinear pooling [28] has recently been used to integrate different CNN features for ﬁne-grained image recognition [17].”
Controllable Invariance through Adversarial Feature Learning (2017, background): “An early work by Tenenbaum and Freeman [1997] propose a bilinear model to learn a representation with the style and content disentangled.”
Multi-content GAN for Few-Shot Font Style Transfer (2017, unknown): “Based on a shape representation, machine learning techniques, including statistical models [27] and bi-linear factorization [30], have been used to infer and transfer stroke styles and composition rules.”
Centering Neural Network Gradient Factors (1996, background): “By combining these two approaches, a test set error of 23% has been reached [27], the lowest we are aware of to date.”
Hierarchical Style-based Networks for Motion Synthesis (2020, methodology): “where B(·) is the bi-linear transformation [29], i.”
Separating Style and Content with Bilinear Models (2000, background): “Preliminary reports of this material appeared in Tenenbaum and Freeman (1997) and Freeman and Tenenbaum (1997).”
Grassmann Pooling as Compact Homogeneous Bilinear Pooling for Fine-Grained Visual Classification (2018, background): “Tenenbaum and Freeman [34] first introduced the bilinear model to separate style and content.”
Neural Networks: Tricks of the Trade (2012, methodology): “Another related architecture is the functional-link network [18], which is similar to a standard MLP but explicitly augments the network with the results of scalar functions applied to the inputs.”
Learning photographic global tonal adjustment with a database of input/output image pairs (2011, unknown): “From this perspective, our dataset offers an opportunity to revisit the style-vs-content problem studied by Tenenbaum and Freeman [17].”
SQA3D: Situated Question Answering in 3D Scenes (2022, methodology): “Earlier endeavors usually comprise multi-modal fusion (Tenenbaum & Freeman, 1996; Perez et al., 2018) and are trained from scratch (Wang et al., 2018; Fried et al., 2018; Wang et al., 2019), while recent efforts would employ pre-trained models (Pashevich et al., 2021; Hong et al., 2021a; Suglia et…”
Feature Extraction Through LOCOCODE (1999, background): “Also, recent experiments (mostly conducted during the time this article has been under review) have shown that even better results can be obtained by using additional context information to improve classification performance (e.g., Turney, 1993; Herrmann, 1997; Tenenbaum & Freeman, 1997 ).”
Image-to-image translation for cross-domain disentanglement (2018, background): “For this reason, there has been a substantial interest on learning disentangled representations [19, 42], including some work based on generative models [25, 33, 36].”
Applications of support vector machines to speech recognition (2004, unknown): “This is worse than the best performance reported on this data set (30% using a speaker adaptation scheme called separable mixture models [29]) but significantly better than the”
Old and New Matrix Algebra Useful for Statistics (2000, methodology): “The differential of an expression can be computed by iteratively applying the following rules: dA = 0 (for constant A) (2) d(αX) = αdX (3) d(X+Y) = dX+ dY (4) d(tr(X)) = tr(dX) (5) d(XY) = (dX)Y +XdY (6) d(X⊗Y) = (dX)⊗Y +X⊗ dY (see section 2) (7) d(X ◦Y) = (dX) ◦Y +X ◦ dY (see section 5) (8) dX = −X”