FishNet: A High-Resolution Image Dataset for Automated Fish Species Recognition
Description
Dataset Overview This dataset comprises 2,455 high-resolution images of commonly found freshwater and brackishwater fish species. The images were captured under natural lighting conditions and are organized into eight folders, each based on a specific species. Class Distribution Mola Carplet (Mola): 405 Swamp Barb (Puti): 495 Mystus Catfish (Tengra): 431 Stinging Catfish (Shing): 549 Prawn: 225 Shrimp: 122 Dwarf Gourami: 228 Data Processing: Original HEIF images were converted to JPEG, resized to 640×640 pixels, and renamed using Python scripts (convert_heif_to_jpg.py, resize_and_rename.py) provided in the code/ folder for reproducibility. Folder Structure & Naming: Images are organized by species folder, with filenames as <FolderName>_<number>.jpg (e.g., Mola fish_1.jpg). Purpose This dataset supports the development of automated fish species classification systems using machine learning and computer vision. It is useful for applications in aquaculture, seafood identification, food quality control, and biodiversity monitoring. It also serves as a valuable resource for training and evaluating image-based recognition models in both research and industry contexts.
Files
Institutions
- Daffodil International University