Ofda-Cnn: A Novel Metaheuristic Algorithm-Based Deep Cnn for Multi-Species Seagrass Classification

Convolutional Neural Network (CNN) based deep learning techniques have become prevalent in the field of image classification in recent years. CNNs have proven to be highly effective in visual recognition tasks, however, CNN-based deep learning model development requires significant architectural engineering and hyperparameter tuning. In this paper, we present a novel metaheuristic algorithm called the Opposition-based Flow Direction Algorithm (OFDA) for tuning the hyperparameters and automating the architecture of CNNs for the use of multi-species seagrass classification tasks. Seagrass species play a crucial role in maintaining biodiversity and stability in marine ecosystems but are threatened by human activities and climate change. Accurate classification of seagrass species is crucial for developing effective management policies. The proposed algorithm is an improved version of the recently proposed Flow Direction Algorithm (FDA) algorithm, which enhances the search capability in the search space by using the Opposition-based learning (OBL) technique. The proposed deep neuroevolution algorithm (OFDA-CNN) is evaluated against eight well-known optimisations algorithm-based CNNs on newly developed multi-species seagrass datasets. This proposed OFDA-CNN model achieves the highest 99.33% overall accuracy among the nine deep neuroevolution algorithms. We also compare the performance to the state-of-the-art peer competitors on the publicly available multi-species seagrass datasets. The proposed model also achieves an overall accuracy of 92.20% and 94.5% on the publicly available four classes and five classes version of the ‘DeepSeagrass’ dataset, respectively and outperforms the state-of-the-art multi-species seagrass classification approaches.