Fish Species Classification Using Machine Learning
Publisher: IEEE
Authors:
B Yalini, Velammal College of Engineering and Technology, MaduraiM Jayameena, Velammal College of Engineering and Technology； Madurai
M Soundarya, Velammal College of Engineering and Technology, Madurai
Abstract:
Fish are the central element of nature and are an important component of mankind. They regulate the structure of the ecosystem, act as a source of food, and are economic and cultural pillars. Taxonomy of fish is defining their species and categorizing them according to their form and action. It is one of the key steps in preserving records of biodiversity, preserving animals that are unique, and in assisting scientists with their work. Our research interest in utilizing Resnet with feature pyramid Network for classification of fishes with images. The architecture captures the finest details and the whole body of the fish as an entirety by a sliding-window and layered framework. Because of that, it performs better, even on near indistinguishable species. Previous fish classification methods were sluggish and were followed by a litany of failures. They included hand tuned feature extraction, provided with little finesse for complicated images, and were unable to distinguish easily between similar organisms. They could not also be scaled for real time utilization, and therefore, they failed without the utilization of deep learning ideas. Our answer addresses these problems by promptly and correctly recognizing fish species from photos. It has the ability to be incorporated in real time use within fisheries, marine research, aquaculture, and conservation work, for improving the overall efficiency and reliability of the process.Fish are the central element of nature and are an important component of mankind. They regulate the structure of the ecosystem, act as a source of food, and are economic and cultural pillars. Taxonomy of fish is defining their species and categorizing them according to their form and action. It is one of the key steps in preserving records of biodiversity, preserving animals that are unique, and in assisting scientists with their work. Our research interest in utilizing Resnet with feature pyramid Network for classification of fishes with images. The architecture captures the finest details and the whole body of the fish as an entirety by a sliding-window and layered framework. Because of that, it performs better, even on near indistinguishable species. Previous fish classification methods were sluggish and were followed by a litany of failures. They included hand tuned feature extraction, provided with little finesse for complicated images, and were unable to distinguish easily between similar organisms. They could not also be scaled for real time utilization, and therefore, they failed without the utilization of deep learning ideas. Our answer addresses these problems by promptly and correctly recognizing fish species from photos. It has the ability to be incorporated in real time use within fisheries, marine research, aquaculture, and conservation work, for improving the overall efficiency and reliability of the process.
