ADINKRA Symbol Recognition Using Classical Machine Learning and Deep Learning

Aug 4, 2026

Artificial intelligence is transforming academia and industry, but black communities and African countries remain underrepresented in AI research and applications. This work promotes accessible AI use by focusing on recognition and classification of traditional Adinkra symbols. The study builds a new ADINKRA image dataset with 174,338 images across 62 symbol classes. A CNN classifier is designed using six convolutional layers and three fully connected layers, and transfer learning with pre-trained VGG and ResNet extracts features for classical models. Performance is evaluated via accuracy and convergence, with visualization of influential prediction regions, forming a baseline for future dataset assessment.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages15

Reading Time38 min

AccessFree to download

Abstract

1 Introduction

Representation and bias in AI

NLP datasets and machine translation

Computer vision and image recognition

Motivating AI in developing countries

ADINKRA SYMBOL RECOGNITION USING CLASSICAL MACHINE LEARNING AND DEEP LEARNING arXiv :2311 . 15728v1 [ cs .CV] 27 Nov 2023 Michael Adjeisah National Centre for Computer Animation Bournemouth University Poole, BH12 5BB, United Kingdom [madjeisah@bournemouth.ac.uk](madjeisah@bournemouth.ac.uk) Kwame Omono Asamoah Zhejiang Normal University Jinhua, 321004, Zhejiang, China [koasamoah2014@gmail.com](koasamoah2014@gmail.com) Martha Asamoah Yeboah College of Computer Science and Technology Zhejiang Normal University Jinhua, Zhejiang, 321004 China [may138@zjnu.edu.cn](may138@zjnu.edu.cn) Raji Rafiu King Provincial Key Laboratory of Electronic, Functional Materials and Devices Huizhou University Huizhou City, 516001, Guangdong Province, China [mrkingraji@outlook.com](mrkingraji@outlook.com) Godwin Ferguson Achaab College of Computer Science and Technology Zhejiang Normal University Jinhua, Zhejiang, 321004 China [achaabf@gmail.com](achaabf@gmail.com) Kingsley Adjei College of Computer Science and Technology Zhejiang Normal University Jinhua, Zhejiang, 321004 China[sleyadjei@gmail.com](sleyadjei@gmail.com) ABSTRACT Artificial intelligence (AI) has emerged as a transformative influence, engendering paradigm shifts in global societies, spanning academia and industry. However, in light of these rapid advances, addressing the underrepresentation of black communities and African countries in AI is crucial. Boosting enthusiasm for AI can be effectively accomplished by showcasing straightforward applications around tasks like identifying and categorizing traditional symbols, such as Adinkra symbols, or familiar objects within the community. In this research endeavor, we dived into classical machine learning and harnessed the power of deep learning models to tackle the intricate task of classifying and recognizing Adinkra symbols. The idea led to a newly constructed ADINKRA dataset comprising 174,338 images meticulously organized into 62 distinct classes, each representing a singular and emblematic symbol. We constructed a CNN model for classification and recognition using six convolutional layers, three fully connected (FC) layers, and optional dropout regularization. The model is a simpler and smaller version of VGG, with fewer layers, smaller channel sizes, and a fixed kernel size. Additionally, we tap into the transfer learning capabilities provided by pre-trained models like VGG and ResNet. These models assist us in both classifying images and extracting features that can be used with classical machine learning models. We assess the model’s performance by measuring its accuracy and convergence rate and visualizing the areas that significantly influence its predictions. These evaluations serve as a foundational benchmark for future assessments of the ADINKRA dataset. We hope this application exemplar inspires ideas on the various uses of AI in organizing our traditional and modern lives. A PREPRINT Keywords Adinkra symbols · Convolutional neural network · Image classification · Machine learning · Transfer learning 1 Introduction AI’s potential for innovation and advancement is undeniable [1], with applications ranging from healthcare [2, 3, 4, 5] and transportation [6, 7] to finance [8, 9] and education [10, 11, 12] . Nevertheless, in these swift developments, it is imperative to prioritize rectifying the inadequate representation of black communities and African nations within AI. The lack of representation raises concerns about the potential bias and limitations that may arise from predominantly homogeneous perspectives, hindering the realization of AI’s full potential and the equitable distribution of its benefits. To comprehensively address this issue, it is necessary to approach it from research areas like Natural Language Processing (NLP) [13] and Computer Vision (CV) [14, 15] perspectives, as they entail the utilization of available datasets and establishing baseline models for condu

FAQs

Why does the paper emphasize representation of black communities and African countries in AI?

It highlights that underrepresentation can create bias and limit AI’s equitable benefits, and it argues for prioritizing this issue while advancing AI research.

What dataset is introduced for Adinkra symbol recognition?

The paper constructs an ADINKRA dataset containing 174,338 images organized into 62 classes, with each class corresponding to a single emblematic Adinkra symbol.

How are classical machine learning and deep learning used together in the study?

A CNN model performs end-to-end classification and recognition, while transfer learning with pre-trained VGG and ResNet extracts image features that can be used with classical machine learning models.

Aug 4, 2026

Uploaded by Ava Thompson