Thermal fish detection system using Deep learning

Abstract

For scholars and image processing professionals, categorization of aquatic organisms in watery imagery is a developing area of study. For fish assessment goals, such as biological auditing ratio, monitoring fish stocks, and preserving critically endangered species, categorization of species of fish in aquatic pictures is crucial. Fish categorization is difficult and time-consuming because of the poor resolution, dark images produced by ocean water's light dispersion and absorption. The excellent fish categorization provided in this work aids the biologist's knowledge of the many species of fish and their environment. For fish categorization in this new methodology, a better deep learning-based automatic encoder decoder strategy was used. Deep learning models usually struggle with the best image classification. The crucial texture featuresare what determine how accurately aquatic species of fish are categorized. The suggested approach is then contrasted with popular deep learning models including Alex-Net, Res Net, VGG Net, and CNN to demonstrate its effectiveness. By using different iterations, groups, and fully associated layers, the proposed technique is examined. Convolutional Neural Network (CNN) is the algorithm that is utilised because it is among the most widely used methods for picture categorization.