SalmonScan: A novel image dataset for machine learning and deep learning analysis in fish disease detection in aquaculture

Fish diseases pose a significant threat to food security in aquaculture, as they can lead to considerable reductions in fish production, quality, and profitability. Globally, salmon aquaculture is the quickest-expanding food production system. Detecting and diagnosing fish diseases in their early stages is essential to prevent the spread of diseases and reduce the negative impact on aquaculture's economy and environment. To serve this purpose, we introduce the SalmonScan dataset, a novel and comprehensive collection of images of healthy and infected salmon fish, which can be used for various applications in computer science and aquaculture. Images from online sources and aquaculture salmon firms were gathered to create the dataset. The dataset was then labeled based on the health status of the fish, fresh or infected. Data augmentation methods like rotation, cropping, flipping, and scaling were used to guarantee the dataset's strength and size. The dataset includes 456 images of fresh fish and 752 images of infected fish, both varied and inclusive while maintaining excellent quality. Other researchers and practitioners can use the dataset we have collected for various purposes. They can use it to create and test new or existing machine learning (ML) and deep learning (DL) based computer vision models for identifying, categorizing, counting, and analyzing the behavior and biomass of salmon fish. They can also use it to study how different environmental factors affect the health and growth of salmon fish. Furthermore, they can evaluate the accuracy and performance of different image acquisition and processing methods. Additionally, they can explore the feasibility of using generative adversarial networks (GANs) and transfer learning to improve the training speed and stability of DL models designed for fish detection. This SalmonScan dataset paper describes and documents the dataset in detail, making it publicly available and reusable for the research community.