Large yellow croaker (Larimichthys crocea) is the largest fish species produced by marine aquaculture in China. Nevertheless, the croaker aquaculture industry is currently confronted with significant challenges, including low feed efficiency (FE), eutrophication in farming areas, and high production costs. In light of these considerations, we conducted a study investigating an intelligent measurement system for individual FE and genomic selection (GS) for FE traits in large yellow croaker. The results demonstrated that: (1) We developed a deep learning-based phenotypic measurement system for automatic real-time measurement of individual feed intake in a group and automatic collection of fish morphometric traits. There were considerable individual differences in FE traits. (2) FE traits in large yellow croaker were a complex trait controlled by multiple micro-effective genes, with a heritability of 0.21. (3) GS for FE in large yellow croaker produced offspring with both improved feed efficiency and growth performance. In addition, the selected lines showed lower ash, crude fat, and crude protein content in feces compared to the control lines. In summary, this study demonstrates the effective intelligent phenotyping method for feed efficiency traits and its application in genomic selection for feed efficiency performance in fish. The approach successfully drives genetic improvement in feed efficiency for large yellow croaker, offering a practical reference for reducing costs, enhancing efficiency, and promoting sustainable aquaculture development.