Fishify: A Mobile-Based Fish Species Identification App with Transfer Learning Using MobileNetV1

Fish fraud is an often-ignored issue which afflicts the seafood industry, threatening both the customer and the ecosystem. A major cause of fish fraud involves mislabeling of seafood species, deliberately providing inaccurate information about the species. It is not only limited to the consumable fish but also extends into the world of aquarium-based fishes where the major issue branches into the origin of the aquarium trade resulting in the disruption of the diversity of aquariums. To address this issue, a CNN model was built, and after hyper-parameter tuning, it was observed that is a need for usage of transfer learning and thus a mobile-based application was developed along with the integration of MobileNetV1, a transfer learning model which facilitates the fish species identification with the help of a single click. The application eases the customer’s uncertainty of a particular fish species by providing an accurate prediction of the fishes.