 Shrimp farming has traditionally been a major source of income for many coastal countries. However, due to the rapid development of society, manual shrimp farms are unable to keep up with the growing demand for faster growth. Automated shrimp farming is therefore necessary to ensure sustainable production. This paper proposes the use of advanced computer vision techniques to improve shrimp farming. Firstly, a high-quality shrimp dataset was established for training deep learning models. Then, a contrastive learning approach was used to reduce the need for large training datasets. The results showed that the proposed method outperformed traditional methods in instance segmentation tasks. Additionally, the concepts discussed in this paper can be applied to other fields that use computer vision technologies. This article was authored by Hangzhou, Sunghoon Kim, Sangchil Kim and others.