Abstract

Image processing is one of the most rapidly evolving technologies today, and it is an approach for applying operations on an image to improve it or extract relevant information from it. This is a critical research field in the engineering and computer sciences. However, analyzing a large number of variables demands a lot of memory and processing resources, which can cause a classification algorithm to overfit the training samples and underfit the test samples. As a result, various strategies, such as extraction, can be used to reduce the number of features in a dataset by producing new features from old ones. In this paper, we first propose a deep learning-based feature extraction approach with a modular neural network, where we employ a pre-trained neural architecture search net (NASNet) as a feature extractor on a custom dataset of raw copper and cobalt images. It allows the input image to be feed-forwarded while performing feature learning and feature mapping and then stops at a pooling layer before the fully connected (FC) layer in the NASNet to extract and save the outputs of that layer in dumped files. Second, the extracted features are used as training data to build a deep neural network and machine learning algorithms for the image classification of copper and cobalt raw minerals. The experimental results show that the NASNet extracts the features efficiently, and the proposed modular neural network performs well with the boosting-decision tree as a classifier, which gives a higher accuracy of 91% than 90% of the deep neural network; moreover, the precision is 1 higher than 0.98 for the deep neural network. Bayesian optimization can be used effectively for hyperparameter tuning in image classification of copper and cobalt raw minerals. When applied to image classification, Bayesian optimization aims to find the best set of hyperparameters for a machine learning model, such as a vision Transformer (ViT) or convolutional neural network (CNN), to achieve optimal classification performance. We used Bayesian optimization on image classification tasks on different datasets such as image classification on copper and cobalt raw minerals image dataset, CIFAR10, CIFAR100, and Satellite image classification Dataset-RSI-CB256. Bayesian optimization is especially valuable in scenarios where evaluating the performance of a model is computationally expensive or time-consuming, such as training deep neural networks on large image datasets. It efficiently explores the hyperparameter space and can significantly improve image classification accuracy. The implementation results of the Bayesian optimization show that the choice of activation function, learning rate, and dropout rate influence the ability of the model to learn and generalize. As a result, the Satellite Image Dataset (RSI-CB256) achieved the highest accuracy of 93% and the lowest loss of 0.3339 among the three datasets. Moreover, CIFAR-10 had a moderate accuracy of 77.5% and a higher loss of 0.9170 compared to CIFAR-100. CIFAR-100 had the lowest accuracy of 53% and the highest loss of 3.1677 among the three datasets. The model performed best on the Satellite Image Dataset (RSI-CB256), indicating that the model and optimized hyperparameters were well-suited for this specific dataset.