A ridgelet neural network approach to sports image classification: Optimization using advanced Parrot algorithm

Recently, the sports images classification has been transformed into an extremely important subject due to its rising applicability in sports-oriented applications. In the present work, a new method is suggested that integrates deep learning and metaheuristic optimization for the categorization of sports images. In exactly that manner, the Ridgelet Neural Network (RNN) is employed for feature learning and classification, and the Advanced Parrot Optimizer (APO) is employed for adjusting the hyperparameters of the RNN to enhance its performance on the selected datasets. Recursive Mean-Separate Histogram Equalization (RMSHE) technique is employed for image preprocessing to improve the contrast and a median filter is employed for noise elimination. Ridgelets being utilized as activation functions enable the RNN to pick features from the input matrix and perform classification processes. The performance of the proposed RNN/APO model is evaluated based on a number of metrics including accuracy, specificity, recall, precision, F1-score, and Matthews Correlation Coefficient (MCC). To validate the efficiency of the provided model, its performance is compared to some of the best models available in the market, i.e., VGG16, Inception V3, Hybridized Hierarchical Deep Convolutional Neural Network (HHDCNN), CNN-TL-DE, and EfficientNetB3/GA. A vast dataset of 6000 frames of videos from some of the most popular sports channels such as YouTube, FIFA, and ESPN is employed to validate the efficiency of the model. The dataset is divided into training (85% data) and testing sets (15% data) to ensure the robustness by sound validation. Experimental results depict the fact that the proposed RNN/APO model performs better than all remaining models with an accuracy rate of 99.61%. The discovery highlights the robustness and potential of the proposed strategy in sport image classification because of its application prospect as a great tool for the field.