Orthogonal Convolutional Neural Networks

Wang, Jiayun; Chen, Yubei; Chakraborty, Rudrasis; Yu, Stella X.

doi:10.1109/cvpr42600.2020.01152

Cited by 132 publications

(128 citation statements)

References 19 publications

Supporting

Mentioning

126

Contrasting

Order By: Relevance

“…Besides, our discriminator, while capable of detecting local artifacts, provides little control for separating such artifacts and global semantic errors. Recent work on image synthesis may provide guidance onto designing better discriminative models [88], training procedures [103], [104], image parametrizations [23], [105], [106] or perceptual loss functions [53]. Second, we proposed a synthesis solution based on manipulating latent spaces within the generative model, but explicitly training the network to generate tileable textures may provide better results than our approach.…”

Section: Discussionmentioning

confidence: 99%

SeamlessGAN: Self-Supervised Synthesis of Tileable Texture Maps

Rodriguez-Pardo,

Garces

2022

Preprint

View full text Add to dashboard Cite

Real-time graphics applications require high-quality textured materials to convey realism in virtual environments. Generating these textures is challenging as they need to be visually realistic, seamlessly tileable, and have a small impact on the memory consumption of the application. For this reason, they are often created manually by skilled artists. In this work, we present SeamlessGAN, a method capable of automatically generating tileable texture maps from a single input exemplar. In contrast to most existing methods, focused solely on solving the synthesis problem, our work tackles both problems, synthesis and tileability, simultaneously. Our key idea is to realize that tiling a latent space within a generative network trained using adversarial expansion techniques produces outputs with continuity at the seam intersection that can then be turned into tileable images by cropping the central area. Since not every value of the latent space is valid to produce high-quality outputs, we leverage the discriminator as a perceptual error metric capable of identifying artifact-free textures during a sampling process. Further, in contrast to previous work on deep texture synthesis, our model is designed and optimized to work with multi-layered texture representations, enabling textures composed of multiple maps such as albedo, normals, etc. We extensively test our design choices for the network architecture, loss function, and sampling parameters. We show qualitatively and quantitatively that our approach outperforms previous methods and works for textures of different types.

show abstract

Section: Discussionmentioning

confidence: 99%

SeamlessGAN: Self-Supervised Synthesis of Tileable Texture Maps

Rodriguez-Pardo,

Garces

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Orthogonalization method using Newton's iteration (ONI) [16] utilizes proxy matrices that can be transformed into orthogonal matrices as the parameters for training and the transformation is conducted through Newton's iteration. Wang et al [40] impose filter orthogonality on a convolutional layer based on the doubly block-Toeplitz matrix representation of the convolutional kernel and derive a new regularization term in the loss function for pursuing orthogonality. These methods can work with existing architectures to improve performance.…”

Section: Related Workmentioning

confidence: 99%

Weight Evolution: Improving Deep Neural Networks Training through Evolving Inferior Weight Values

Lin

Guo

Xing

et al. 2021

Proceedings of the 29th ACM International Conference on Multimedia

View full text Add to dashboard Cite

To obtain good performance, convolutional neural networks are usually over-parameterized. This phenomenon has stimulated two interesting topics: pruning the unimportant weights for compression and reactivating the unimportant weights to make full use of network capability. However, current weight reactivation methods usually reactivate the entire filters, which may not be precise enough. Looking back in history, the prosperity of filter pruning is mainly due to its friendliness to hardware implementation, but pruning at a finer structure level, i.e., weight elements, usually leads to better network performance. We study the problem of weight element reactivation in this paper. Motivated by evolution, we select the unimportant filters and update their unimportant elements by combining them with the important elements of important filters, just like gene crossover to produce better offspring, and the proposed method is called weight evolution (WE). WE is mainly composed of four strategies. We propose a global selection strategy and a local selection strategy and combine them to locate the unimportant filters. A forward matching strategy is proposed to find the matched important filters and a crossover strategy is proposed to utilize the important elements of the important filters for updating unimportant filters. WE is plug-in to existing network architectures. Comprehensive experiments show that WE outperforms the other reactivation methods and plug-in training methods with typical convolutional neural networks, especially lightweight networks. Our code is available at https://github.com/BZQLin/Weight-evolution. CCS CONCEPTS• Computing methodologies → Neural networks.

show abstract

“…W ∈ R D ×D is the weight matrix of the layer and x ∈ R D is the output of the preceding layer. The linear operation in a convolutional layer may be represented as a doubly-block Toeplitz matrix [14]. Another way to perform the operation is to employ reshaping operators to represent the linear operator as a dense matrix applied to all the patches extracted from the input [15].…”

Section: A Weight Matrix As a Product Of Sparse Matricesmentioning

confidence: 99%

PSM-nets: Compressing Neural Networks with Product of Sparse Matrices

Giffon

Ayache

Kadri

et al. 2021

2021 International Joint Conference on Neural Networks (IJCNN)

View full text Add to dashboard Cite

Over-parameterization of neural networks is a well known issue that comes along with their great performance. Among the many approaches proposed to tackle this problem, low-rank tensor decompositions are largely investigated to compress deep neural networks. Such techniques rely on a low-rank assumption of the layer weight tensors that does not always hold in practice. Following this observation, this paper studies sparsity inducing techniques to build new sparse matrix product layers for high-rate neural networks compression. Specifically, we explore recent advances in sparse optimization to replace each layer's weight matrix, either convolutional or fully connected, by a product of sparse matrices. Our experiments validate that our approach provides a better compression-accuracy trade-off than most popular low-rank-based compression techniques.

show abstract

Orthogonal Convolutional Neural Networks

Cited by 132 publications

References 19 publications

SeamlessGAN: Self-Supervised Synthesis of Tileable Texture Maps

SeamlessGAN: Self-Supervised Synthesis of Tileable Texture Maps

Weight Evolution: Improving Deep Neural Networks Training through Evolving Inferior Weight Values

PSM-nets: Compressing Neural Networks with Product of Sparse Matrices

Contact Info

Product

Resources

About