Recent Papers

Marine species classification is one such critical endeavor spanning across biodiversity conservation and ecological studies, fully embracing the current high urgency in establishing useful instruments for the constant surveillance of ocean ecosystems. Unfortunately, traditional classification methods tend to fall short due to factors such as varied underwater conditions or light and limited annotated data. Thus, this research proposes a new hybrid approach, mixing deep learning with machine learning to improve classification accuracy. Basically, a pre-trained ResNet model is set as the feature extraction technique, then followed in turn by RF and KNN classification methods. The dataset incorporates the representation of three marine speciesjellyfish, otters, and sharksand its augmentation is applied as an advanced technique in order to be robust. The hybrid methodology applies the meaningful feature mapping with its ResNet18 and ResNet50 and is offered for prediction using RF and KNN classifiers. Though huge experimentation clearly suggested an accuracy value of 95.97% by the Random Forest classifier and 98.66% by the KNN classifier. Comprehensive evaluation metrics, including confusion matrices, precision, recall, and F1-score, have shown a balanced performance of the models across all classes. Finally, the visualization of predicted and actual classifications provides insight into the model's reliability. This research demonstrates the efficiency of combining deep-learning-based feature extraction with machine learning classifiers to classify marine species, building an enormous foundation for hybrid models implemented in ecology work, thus allowing sizeable and accurate biodiversity remediation systems. For the next task, this method will be extended by larger datasets, alongside the construction and integration of other classifiers for better performance.

Copyright © 2025 Amit Kumar Pandey. This is an open access article distributed under the Creative Commons Attribution License.