Computer vision model with novel cuckoo search based deep learning approach for classification of fish image

• Published in: Multimedia tools and applications 2023, Vol.82 (3), p.3677-3696
• Main Authors: Aziz, Rabia Musheer, Desai, Nishq Poorav, Baluch, Mohammed Farhan
• Format: Article
• Language: English
• Subjects: Artificial neural networks; Classification; Computer Communication Networks; Computer Science; Computer vision; Data search; Data Structures and Information Theory; Deep learning; Fish; Genetic algorithms; Image classification; Machine learning; Multimedia Information Systems; Neural networks; Optimization; Optimization techniques; Search algorithms; Special Purpose and Application-Based Systems
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-022-13437-3

Fish is one of the most important cold-blooded animal groups. Fish is an important part of a healthy diet since it contains several minerals and micronutrients that are necessary for general body development. Because different kinds of fish have varied symptoms when it comes to sickness and decay, it’s critical that we be able to identify and classify the most essential fish species. Traditional methods in this domain are now tedious and slow, however systems based on better deep learning techniques can overcome them. This study proposed a Deep Learning Artificial Neural Network (DLANN) model with a novel optimization technique for fish image classification. The success of DLANN is primarily attributed to its architecture, the optimization technique used, and the tuning of hyperparameters to identify different patterns in data. The Cuckoo Search (CS) algorithm is a popular nature-inspired optimization technique used to solve real-time science and engineering problems. In this paper, to overcome the shortcoming of CS by introducing a Genetic Algorithm (GA) in the exploration phase of the CS approach. A new optimization technique (GA-CS) has been proposed for DLANN to solve problems in fish image classification. An extensive experiment was conducted to compare the performance of the proposed techniques with several popular (EfficientNet, Inception V3, ResNet150 V2, VGG-19, DenseNet 121, LSTM Model, and a personalized Convolutional Neural Network (CNN) model) techniques of deep learning. Experimental results with different evaluation matrices (classification accuracy, recall, precision, standard deviation, and F1- Scores) show that the proposed optimization technique with deep learning gives the best result for fish image classification.