Instance Segmentation of Shrimp Based on Contrastive Learning

Shrimp farming has traditionally served as a crucial source of seafood and revenue for coastal countries. However, with the rapid development society, conventional small-scale manual shrimp can no longer meet increasing demand growth. As result, it is imperative to continuously develop automation technology efficient large-scale farming. Smart represents an innovative application advanced technologies management practices in aquaculture expand scale production. Nonetheless, use these new not without difficulties, including scarcity public datasets high cost labeling. In this paper, we focus on computer vision techniques To achieve objective, first establish high-quality dataset training various deep learning models. Subsequently, propose method that combines unsupervised downstream instance segmentation tasks mitigate reliance large datasets. Our experiments demonstrate involving contrastive outperforms direct fine-tuning model tasks. Furthermore, concepts presented paper extend other fields utilize technologies.