Underwater Animal Identification and Classification Using a Hybrid Classical-Quantum Algorithm
Sheena Christabel Pravin, G. Rohith, V. Kiruthika, E. Manikandan, S. Methelesh, Akash L. Manoj
Abstract
Underwater Animal Identification and Classification is gaining significance in recent times as a result of the increased need for biodiversity monitoring and ecological surveillance. Although classical deep learning techniques have been widely used for these tasks, the effectiveness of identification and classification remains a bottleneck for several researchers due to the live capture of animals in complex environments, a limited sea-animal image dataset, and the complex topography of the seafloor, particularly in shallow waters, sediments, reefs, submarine ridges, and ship radiation. In this paper, three hybrid Classical-Quantum neural networks ResNet50-QCNN, ResNet18-QCNN and InceptionV3-QCNN have been proposed for underwater quantum-classical Animal Identification and Classification. By using quantum devices to reduce dimension and denoise datasets, it dramatically reduces the complexity of data processing on classical computers. The outcomes of the numerical simulation show that the quantum algorithm may effectively reduce dimensionality and increase classification accuracy. Even when quantum data is read out classically, the hybrid approach provides polynomial acceleration in dimension reduction beyond classical procedures. The three hybrid models, viz., ResNet50-QCNN, ResNet18-QCNN, and InceptionV3-QCNN, displayed classification test accuracy of 88%, 80.29%, and 70%, respectively, revealing that ResNet50-QCNN performed best in identifying and classifying underwater animals.