BDFreshFish: A Comprehensive Image Dataset of Bangladeshi Freshwater Fishes to Develop AI-driven Solutions for Autonomous Fisheries Management

The BDFreshFish dataset that is presented in the paper contains image data that corresponds to eight different types of native freshwater fish that are found in various parts of Bangladesh. The following categories are distinguished by their scientific names: Heteropneustes Fossilis, Mastacembelus Armatus, Ompok Bimaculatus, Channa Punctata, Marcrobrachium Malcoimsonii, Anabas Testudineus, and Batasio Tengana. Across these categories, the dataset includes over 3100 images, with 300 augmented images and roughly 100 raw images per class. Bangladeshi rivers provided the source of most of the raw photos. The head, body, scales, and fins are the four main features of each fish species, highlighted in each image in the collection. The images were taken against appropriate backdrops under ideal natural lighting. The collection was well-structured, with every picture in its proper folder, making it easy for different deep-learning and machine-learning models to use. Researchers can use deep learning (DL) and traditional machine learning (TML) techniques with this large dataset. Scientists can advance their knowledge of the freshwater ecosystem and fish migration by examining distinct fish category classifications.