Color Equivariant

Color equivariance in neural networks focuses on designing models that consistently respond to variations in color (hue and saturation) while preserving relevant information. Current research explores convolutional neural networks (CNNs) and generative adversarial networks (GANs), incorporating concepts from group theory and partial differential equations to achieve this equivariance, often through specialized convolutional layers or morphological operations. This research aims to improve the robustness and generalizability of models to color variations in image data, leading to more accurate and reliable performance in computer vision tasks and potentially impacting applications like object recognition and image generation. The development of hyperparameter-agnostic optimization methods further enhances the practicality of these color-equivariant models.

Papers

June 13, 2024

Learning Color Equivariant Representations
Felix O'Mahony, Yulong Yang, Christine Allen-Blanchette
Convolutional Neural Network Equivariant Network Equivariant Convolutional Neural Network Perceptual Adjustment Query Color Equivariant

March 22, 2024

Geometric Generative Models based on Morphological Equivariant PDEs and GANs
El Hadji S. Diop, Thierno Fall, Alioune Mbengue, Mohamed Daoudi
Generative Adversarial Network Single GAN Morphological Symmetry Geometric Generative Model Color Equivariant PDE G CNNs

October 30, 2023

Color Equivariant Convolutional Networks
Attila Lengyel, Ombretta Strafforello, Robert-Jan Bruintjes, Alexander Gielisse, Jan van Gemert
Convolutional Neural Network Inter Part Equivariance Color Constancy Color Equivariant

October 21, 2023

Towards Hyperparameter-Agnostic DNN Training via Dynamical System Insights
Carmel Fiscko, Aayushya Agarwal, Yihan Ruan, Soummya Kar, Larry Pileggi, Bruno Sinopoli
Deep Neural Network Dynamical System DNN Model Hyperparameter Tuning Model Agnostic Training Color Equivariant

August 15, 2022

Rethinking Graph Neural Networks for the Graph Coloring Problem
Wei Li, Ruxuan Li, Yuzhe Ma, Siu On Chan, David Pan, Bei Yu
State of the Art GNN Model Graph Coloring Color Equivariant

Color Equivariant

Papers

Learning Color Equivariant Representations

Geometric Generative Models based on Morphological Equivariant PDEs and GANs

Color Equivariant Convolutional Networks

Towards Hyperparameter-Agnostic DNN Training via Dynamical System Insights

Rethinking Graph Neural Networks for the Graph Coloring Problem