Texture Synthesis Using Shallow Convolutional Networks with Random Filters

Here we demonstrate that the feature space of random shallow convolutional neural networks (CNNs) can serve as a surprisingly good model of natural textures. Patches from the same texture are consistently classified as being more similar then patches from different textures. Samples synthesized from the model capture spatial correlations on scales much larger then the receptive field size, and sometimes even rival or surpass the perceptual quality of state of the art texture models (but show less variability). The current state of the art in parametric texture synthesis relies on the multi-layer feature space of deep CNNs that were trained on natural images. Our finding suggests that such optimized multi-layer feature spaces are not imperative for texture modeling. Instead, much simpler shallow and convolutional networks can serve as the basis for novel texture synthesis algorithms.

• Publication year

  2016

• Venue

  arXiv.org

• Publication date

  2016-05-31

• Fields of study

  Computer Science

• Identifiers
  arXiv 1606.00021
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Theano: A Python framework for fast computation of mathematical expressions

  2016cited by this paper
• Lasagne: First release.

  2015cited by this paper
• Texture Synthesis Using Convolutional Neural Networks

  2015influential reference
• Very Deep Convolutional Networks for Large-Scale Image Recognition

  2014influential reference
• ImageNet Large Scale Visual Recognition Challenge

  2014cited by this paper
• Audio Texture Synthesis with Scattering Moments

  2013cited by this paper
• Understanding the difficulty of training deep feedforward neural networks

  2010cited by this paper
• Efficient Implementation of the K-SVD Algorithm using Batch Orthogonal Matching Pursuit

  2008cited by this paper
• SciPy: Open Source Scientific Tools for Python

  2001cited by this paper
• Exploring Texture Ensembles by Efficient Markov Chain Monte Carlo-Toward a 'Trichromacy' Theory of Texture

  2000cited by this paper
• A Parametric Texture Model Based on Joint Statistics of Complex Wavelet Coefficients

  2000influential reference
• Minimax Entropy Principle and Its Application to Texture Modeling

  1997cited by this paper
• On the rationale of maximum-entropy methods

  1982cited by this paper
• Visual Pattern Discrimination

  1962cited by this paper

• Orientation-Aware Texture Synthesis Using Rotated Texture Samples With CNNs

  2025cites this paper
• Multispectral Texture Synthesis Using RGB Convolutional Neural Networks

  2024cites this paper
• On Synthetic Texture Datasets: Challenges, Creation, and Curation

  2024cites this paper
• Texture-like representation of objects in human visual cortex

  2022cites this paper
• Neural Style Transfer with distortion handling for Audio and Image

  2022cites this paper
• Efficient processing of natural scenes in visual cortex

  2022cites this paper
• StyleTime: Style Transfer for Synthetic Time Series Generation

  2022cites this paper
• Landscape Similarity Analysis Using Texture Encoded Deep-Learning Features on Unclassified Remote Sensing Imagery

  2021cites this paper
• Bi-encoder Network with Structure-texture Consistency for Image Inpainting

  2021cites this paper
• Environment Transfer for Distributed Systems

  2021cites this paper
• Optimal Textures: Fast and Robust Texture Synthesis and Style Transfer through Optimal Transport

  2020cites this paper
• Audio style transfer using shallow convolutional networks and random filters

  2020cites this paper
• How big can style be? Addressing high dimensionality for recommending with style

  2019cites this paper
• Scale Space and Variational Methods in Computer Vision

  2019cites this paper
• Applying visual domain style transfer and texture synthesis techniques to audio: insights and challenges

  2019cites this paper
• Landcover classification using texture-encoded convolutional neural networks : peeking inside the black box

  2019cites this paper
• Macrocanonical Models for Texture Synthesis

  2019cites this paper
• A Bayesian Perspective on the Deep Image Prior

  2019cites this paper
• Audio Texture Synthesis with Random Neural Networks: Improving Diversity and Quality

  2019cites this paper
• Sound texture synthesis using convolutional neural networks

  2019cites this paper
• Mechanisms of Artistic Creativity in Deep Learning Neural Networks

  2019cites this paper
• Thoracic MRI Emulation Through Texture Synthesis and Ground Truth Manipulation

  2019cites this paper
• Sound Texture Synthesis Using RI Spectrograms

  2019cites this paper
• Maximum entropy methods for texture synthesis: theory and practice

  2019cites this paper
• Texture synthesis with neural networks

  2018cites this paper
• Deep Learning for Audio Style Transfer

  2018cites this paper
• Deep learning framework for infiltrative lung disease classification

  2018cites this paper
• Bilinear Convolutional Neural Networks for Fine-Grained Visual Recognition

  2018cites this paper
• Innovative Non-parametric Texture Synthesis via Patch Permutations

  2018cites this paper
• Synthesizing Diverse, High-Quality Audio Textures

  2018cites this paper
• Adversarial Reprogramming of Neural Networks

  2018cites this paper
• Deep Learning Methods for Texture Analysis in Medical Imaging

  2018cites this paper
• Musical Synthesis for Certain Music Styles Based on Machine Learning Algorithms

  2018cites this paper
• Transformation-Grounded Image Generation Network for Novel 3D View Synthesis

  2017cites this paper
• Audio Style Transfer

  2017cites this paper
• Cascade of convolutional neural networks for lung texture classification: overcoming ontological overlapping

  2017cites this paper
• Unifying local and non-local signal processing with graph CNNs

  2017cites this paper
• Visual attribute transfer through deep image analogy

  2017cites this paper
• Texture and art with deep neural networks.

  2017cites this paper
• A parametric texture model based on deep convolutional features closely matches texture appearance for humans

  2017influential citation
• On random weights for texture generation in one layer CNNS

  2016influential citation
• Incorporating long-range consistency in CNN-based texture generation

  2016cites this paper
• Texture synthesis through convolutional neural networks and spectrum constraints

  2016cites this paper
• Authoring image decompositions with generative models

  2016cites this paper
• Stationary dynamic texture synthesis using convolutional neural networks

  2016cites this paper
• On Random Weights for Texture Generation in One Layer Neural Networks

  2016cites this paper
• Inferring low-dimensional microstructure representations using convolutional neural networks

  2016influential citation
• Bilinear CNNs for Fine-grained Visual Recognition

  2015cites this paper
• Bilinear CNN Models for Fine-Grained Visual Recognition

  2015cites this paper
• Cross-Convolutional-Layer Pooling for Image Recognition

  2015cites this paper