Hexagonal Image Processing in the Context of Machine Learning - Conception of a Biologically Inspired Hexagonal Deep Learning Framework

Aug 5, 2026

Upload by Ophelia

Hexagonal image processing in machine learning is developed by drawing inspiration from human visual perception. The work proposes image processing systems that combine biologically motivated evolutionary structures, aiming to leverage advantages over conventional square-lattice approaches used by many recording and output devices. It introduces Hexnet, the hexagonal transformation pipeline, and dependent methods. Experiments on a test environment show improved hexagonal image processing performance, including benefits for hexagonal artificial neural networks (H-DNN), with reduced trainable parameters and higher training and test rates.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages8

Reading Time20 min

AccessFree to download

Introduction and Motivation

Related Work

Hexagonal Image Processing in the Context of Machine Learning: Conception of a Biologically Inspired Hexagonal Deep Learning Framework Tobias Schlosser, Michael Friedrich, and Danny Kowerko Junior Professorship of Media Computing, Chemnitz University of Technology, 09107 Chemnitz, Germany, { [firstname.lastname}@cs.tu-chemnitz.de](firstname.lastname}@cs.tu-chemnitz.de) arXiv : 19 11 . 11251v3 [ cs .LG] 26 Jan 2020 Abstract—Inspired by the human visual perception system, hexagonal image processing in the context of machine learning deals with the development of image processing systems that combine the advantages of evolutionary motivated structures based on biological models. While conventional state of the art image processing systems of recording and output devices almost exclusively utilize square arranged methods, their hexagonal counterparts offer a number of key advantages that can beneﬁt both researchers and users. This contribution serves as a general application-oriented approach with the synthesis of the therefor designed hexagonal image processing framework, called Hexnet, the processing steps of hexagonal image transformation, and dependent methods. The results of our created test environment show that the realized framework surpasses current approaches of hexagonal image processing systems, while hexagonal artiﬁcial neural networks can beneﬁt from the implemented hexagonal architecture. As hexagonal lattice format based deep neural networks, also called H-DNN, can be compared to their square counterpart by transforming classical square lattice based datasets into their hexagonal representation, they can also result in a reduction of trainable parameters as well as result in increased training and test rates. Index Terms—Computer Vision, Pattern and Image Recognition, Deep Learning, Convolutional Neural Networks, Hexagonal Image Processing, Hexagonal Lattice, Hexagonal Sampling I. INTRODUCTION AND MOTIVATION Following the developments of recent years, machine learning methods in the form of artiﬁcial neural networks are becoming increasingly important. Convolutional neural networks (CNN) for object recognition and classiﬁcation in this context are one of the approaches that are in the focus of current research in the domain of deep learning. In order to meet the ever-increasing demands of more complex problems, novel application areas, and larger data sets, following Krizhevsky et al. [1], more and more novel models and procedures are being developed, as their complexity and diversity following Szegedyet al. [2] steadily increases. However, novel architectures for artiﬁcial neural networks, including convolutional and pooling layers, are being developed that exceed the conventional cartesian-based architectures. These include, but are not limited to, spherical or non-Euclidean manifolds, as current research by Bronstein et al. [3] demonstrates. While the structure and functionality of artiﬁcial neural networks is inspired by biological processes, they are also limited by their underlying structure due to the current state of the art of recording and output devices. Therefore, mostly square structures are used, which also signiﬁcantly restrict subsequent image processing systems, as the set of allowed operations is reduced depending on the arrangement of the underlying structure [4] . In comparison to the current state of the art in machine learning, the human visual perception system suggests an alternative, evolutionarily-based structure, which manifests itself in the human eye. The retina of the human eye displays according to Curcio et al. [5] a particularly strong hexagonal arrangement of sensory cells, whereas following Middleton and Sivaswamy [6] the processing of incoming signals is handled much more efﬁciently. Through the layers of the retina the reduction of the received information to the nerve ﬁbers takes place, which are connected to the brain via afferents over the optic nerve.

FAQs

What problem does hexagonal image processing address compared with square-based methods?

Square-based image processing systems restrict subsequent operations due to the underlying cartesian structure. Hexagonal lattice formats offer a more suitable alternative inspired by biological vision and can improve transformation efficiency and related processing outcomes.

What is Hexnet in the proposed framework?

Hexnet is the designed hexagonal image processing framework that synthesizes the processing steps of hexagonal image transformation and related dependent methods for machine learning contexts.

How do hexagonal neural networks (H-DNN) relate to conventional square-lattice datasets?

Classical square-lattice datasets can be transformed into hexagonal representations, enabling comparison between hexagonal and square architectures while potentially reducing trainable parameters and improving training and test rates.

Aug 5, 2026

Upload by Ophelia