Optimizing Image Classification Through Preprocessing and Feature Engineering Techniques

Abstract

A collection of labelled photos obtained from several sources (such as repositories and sensors) is used in this investigation. In order to assess the efficacy of the model, the dataset is partitioned into three parts: training, validation, and test. The data is improved and the model's resilience is increased by applying image preprocessing techniques including scaling, normalization, and augmentation. Three machine learning techniques are investigated: Random Forest Classifier, k-Nearest Neighbors (kNN), and Convolutional Neural Networks: CNN. The CNN model gets an impressive test accuracy of 84.39% and a training accuracy of 91.12%. After evaluating each model's performance using accuracy, precision, recall, and confusion matrices, the kNN classifier achieves 63.08% accuracy while the Random Forest Classifier achieves 69.83% accuracy. The results of this study provide light on how various models do when faced with picture categorization problems.