The FlatIMG dataset is the first publicly available, high-quality image dataset specifically curated for external lesion detection in flatfish. It consists of 1,861 high-resolution images of 164 diseased flatfish collected from 10 aquaculture farms in South Korea. A total of 394 lesion cases are annotated by aquatic disease experts, covering a wide range of lesion types such as ulceration, hemorrhage, lymphocystis, erosion, and prolapse.

Each image is annotated with detailed information, including the lesion’s anatomical location (head, fins, or body), side (ocular or blind), and, when applicable, the associated pathogen (bacteria, viruses, or parasites). The fish were photographed outside water using a 2048×1536 resolution camera to capture clear external symptoms. Heatmaps are also provided to visualize lesion occurrence patterns across the fish body. This dataset enables robust development of lesion detection models and supports research in aquaculture health monitoring and fish pathology.