What question did this study set out to answer?

To enhance the classification accuracy of fish species using advanced deep learning techniques.

February 27, 2026Open Access

Improved Classification Model for Fish Species Based on Deep Learning Techniques

Key Points

To enhance the classification accuracy of fish species using advanced deep learning techniques.
Utilized Convolutional Neural Networks (CNNs) including VGG16 and ResNet50.
Applied transfer learning and fine-tuning strategies for optimization.
Conducted image preprocessing and augmentation such as rotation, scaling, and normalization.
Trained, validated, and tested models on a Fish Image Dataset from kaggle.com.
Achieved overall accuracy of 96.2% with the improved ResNet50 model.
Precision of 97.4%, recall of 96.2%, and F1-score of 96.4% were reported.
Improved model outperformed baseline CNN and VGG16 models.

Abstract

Abstract: Accurate classification of fish species plays a crucial role in biodiversity conservation, fisheries management, and aquatic ecosystem monitoring. Traditionally, fish classification methods, rely on manual identification, which are time-consuming, error-prone, and often depend heavily on expert knowledge. To overcome these limitations, this study presents an Improved Classification Model for Fish Species Based on Deep Learning Techniques. The proposed model integrates Convolutional Neural Networks (CNNs) with advanced deep learning architectures such as VGG16, ResNet50 and optimized through transfer learning and fine-tuning strategies to enhance feature extraction and classification accuracy. Fish Image Dataset containing multiple species acquired from kaggle.com was utilized. The dataset was preprocessed using image augmentation techniques including rotation, scaling, flipping, and brightness adjustments. Images were normalized and resized to 224×224 pixels to fit the model input requirements. The study adopted a comparative experimental design where three CNN architectures were trained, validated, and tested on the same dataset. Each model was evaluated using accuracy, precision, recall and F1-score metrics to ensure robust performance assessment. Training was conducted using Python and TensorFlow/Keras frameworks. The experimental results revealed that the Improved ResNet50-based model achieved superior classification performance compared to baseline CNN and VGG16 models, attaining an overall accuracy of 96.2%, precision of 97.4%, recall of 96.2%, and an F1-score of 96.4% on the test set. The improved deep learning model provides a robust, scalable, and automated approach for fish species classification, offering substantial potential for applications in marine biology, aquaculture, and ecological research. Keywords: Fish Image Recognition, Classification, CNN, VGG16, ResNet, Deep Learning. Title: Improved Classification Model for Fish Species Based on Deep Learning Techniques Author: Samuel Oluwatayo Ogunlana, Felix Ola Aranuwa International Journal of Computer Science and Information Technology Research ISSN 2348-1196 (print), ISSN 2348-120X (online) Vol. 14, Issue 1, January 2026 - March 2026 Page No: 110-121 Research Publish Journals Website: www.researchpublish.com Published Date: 21-February-2026 DOI: https://doi.org/10.5281/zenodo.18723919 Paper Download Link (Source) https://www.researchpublish.com/papers/improved-classification-model-for-fish-species-based-on-deep-learning-techniques

Improved Classification Model for Fish Species Based on Deep Learning Techniques

Key Points

Abstract

Cite This Study