Datasets Comparison

MatsyaDx-BD is an image dataset of freshwater fish diseases collected from aquaculture farms in Rajshahi Division, Bangladesh. The dataset contains 2137 RGB images representing three commonly farmed freshwater fish species: Grass Carp (Ctenopharyngodon idella), Rohu (Labeo rohita), and Silver Carp (Hypophthalmichthys molitrix). The images cover four health conditions observed in freshwater aquaculture systems: Bacterial Gill Disease (275 images), Bacterial Red Disease (653 images), Epizootic Ulcerative Syndrome (EUS, 332 images), and Healthy Fish (877 images). Fish body length ranges are: Grass Carp 20–44 cm, Rohu 23–40 cm, and Silver Carp 32–50 cm. All images were captured using a Samsung Galaxy S25 Ultra smartphone under natural aquaculture farm conditions, resulting in variations in lighting, background, and fish orientation. The dataset is organized in a class-wise folder structure, where images are grouped according to fish health condition. It can support AI, computer vision, and IoT applications for automated disease detection, as well as aquaculture research and studies of regional disease patterns. The dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license, allowing reuse with appropriate attribution.

Collect images of freshwater carp species (grass carp, rohu, and silver carp) from aquaculture farms or ponds under natural lighting conditions using a smartphone or camera. Capture multiple images from different angles for both diseased and healthy fish. Remove low-quality images and manually label them according to visible disease symptoms (bacterial gill disease, bacterial red disease, EUS, and healthy fish). Finally, organize the images into class-labeled folders for dataset preparation.

Computer Science, Computer Vision, Machine Learning, Aquaculture Disease, Deep Learning, Freshwater Aquaculture, IoT Application, Aquaculture Farm

Creative Commons Attribution 4.0 International

MatsyaDx-BD is a curated image dataset of freshwater fish diseases collected from aquaculture farms in the Rajshahi Division, Bangladesh. The dataset contains a total of 2137 RGB images representing three widely farmed freshwater carp species: Grass Carp (Ctenopharyngodon idella), Rohu (Labeo rohita), and Silver Carp (Hypophthalmichthys molitrix). These images cover four major health conditions commonly observed in freshwater aquaculture systems, including Bacterial Gill Disease (275 images), Bacterial Red Disease (653 images), Epizootic Ulcerative Syndrome (EUS) (332 images), and Healthy Fish (877 images). The fish body length ranges are 20–44 cm for Grass Carp, 23–40 cm for Rohu, and 32–50 cm for Silver Carp. All images were captured using a Samsung Galaxy S25 Ultra smartphone under natural aquaculture farm conditions, resulting in realistic variations in lighting, background, and fish orientation. In this updated version, several improvements have been introduced to enhance the dataset’s usability, structure, and reproducibility. The dataset has been reorganized into a specimen-wise folder structure consisting of 86 distinct specimens, which enables better traceability and supports fine-grained analysis at the individual specimen level. Additionally, the metadata has been improved and enriched to provide clearer annotations and facilitate easier integration into machine learning and computer vision workflows. All images have been uniformly resized to a resolution of 4000 × 3000 pixels to ensure consistency in input dimensions across different models and experiments. Importantly, no changes have been made to the original raw data content, thereby preserving the authenticity and integrity of the dataset. The dataset is organized in a hierarchical structure where images are grouped by individual specimens and annotated with corresponding disease labels and species information. This enhanced organization supports both specimen-level analysis and traditional class-wise disease classification tasks, making the dataset suitable for applications such as deep learning-based fish disease detection, computer vision research in aquaculture, IoT-based smart aquaculture monitoring systems, and regional disease pattern analysis. The dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license, allowing reuse and distribution with proper attribution.

To reproduce a similar dataset, images of freshwater carp species such as Grass Carp, Rohu, and Silver Carp should be collected from aquaculture farms or ponds under natural lighting conditions using a smartphone or camera. Multiple images should be captured from different angles for both diseased and healthy fish to ensure diversity in the dataset. After data collection, low-quality or blurred images should be removed through a quality control process. The remaining images should then be manually annotated based on visible disease symptoms, including Bacterial Gill Disease, Bacterial Red Disease, Epizootic Ulcerative Syndrome (EUS), and healthy conditions. Finally, the dataset should be organized into a structured format, preferably using a specimen-wise or class-wise folder hierarchy, along with proper metadata annotations. For improved consistency in machine learning applications, image resolution can be standardized, such as resizing all images to a uniform dimension.

Computer Science, Computer Vision, Machine Learning, Aquaculture Disease, Deep Learning, Freshwater Aquaculture, IoT Application, Aquaculture Farm

Creative Commons Attribution 4.0 International