<p>Image classification is one of the most popular fields in machine learning. The ability to classify depends largely on obtaining good features from images, and if the features used are of superior quality, they lead to high-level classification accuracy. Although significant progress has been witnessed in machine learning (ML) and deep-learning, which covers areas such as image identification, the problems remain to be solved. Some of the complexities may reduce the accuracy of classification into specific classes. Advocating challenging the matching process, researchers suggest a new approach to improve image classification accuracy. This means the process of using the significant features from the already known neural networks, for example, ResNet-18 and ResNet-50, and coming up with a hybrid feature extraction system that utilizes the binary marine predator algorithm (BMPA)as well as the gray wolf optimization (GWO) algorithm. Then, through a combination named BMPA_GWO, the results of the said algorithm are applied on three different image datasets. The performance of said classifier, support vector machine (SVM), is the evaluator. The experimental results show the evident advantage of the project method compared to the conventional methods. The study underscores the inadequacy of relying solely on a single feature extractor, shallow or deep. Instead, an integrated approach combining deep learning methodologies with features from the hybrid algorithm proves more effective for Image classification tasks.</p>