Comparative Analysis of Deep Learning Models for Marine Species Classification: CNN, ResNet50, and SE-CNN Approaches |
Author(s): |
| Prashant Yadav , Thakur College of Science and Commerce; Prashant Yadav, Thakur College of Science and Commerce; Aman Mishra , Thakur College of Science and Commerce; Amit Pandey , Thakur College of Science and Commerce; Santosh Kumar Singh, Thakur College of Science and Commerce |
Keywords: |
| Convolutional Neural Network (CNN), Squeeze-and-Excitation CNN (SE-CNN), Deep Learning Models, ResNet50 |
Abstract |
|
The classification of marine species is essential for biodiversity conservation and ecological monitoring. In this study, we compare three deep learning models—Standard Convolutional Neural Network (CNN), Transfer Learning using ResNet50, and Squeeze-and-Excitation CNN (SE-CNN)—to evaluate their effectiveness in identifying marine species. A dataset consisting of 16,616 images of six marine species was used for training and testing. The models were assessed based on accuracy, computational efficiency, and class-wise performance analysis. Experimental results indicate that SE-CNN outperforms the traditional CNN model, while ResNet50 achieves high accuracy with minimal training effort. Our findings provide insights into selecting optimal models for marine species classification and contribute to AI-driven ecological monitoring solutions. |
Other Details |
|
Paper ID: IJSRDV13I10057 Published in: Volume : 13, Issue : 1 Publication Date: 01/04/2025 Page(s): 97-98 |
Article Preview |
|
|
|
|
