Answers are grounded in this paper and cite their sources.
Molecular breeding has advanced genetic improvement in aquaculture, yet efficient and accurate acquisition of large-scale phenotypic data remains a key limitation due to the absence of intelligent, high-throughput solutions. However, there is no suitable model specifically designed for extracting comprehensive phenomic data in golden pompano ( Trachinotus ovatus ). To address this gap, TovNet-Vision, a deep learning framework for external phenotypic analysis, was developed. At the heart of this framework is the Lightweight and Efficient Convolution (LEC) module, which was specifically designed to maximize feature extraction while minimizing model parameters, making it well-suited for deployment on resource-constrained devices. After incorporating the LEC module, performance (mAP @50M ) improved by 5.1% compared to the baseline model. The LEC module, together with the Lightweight shared convolutional separator batch-normalization detection head, was integrated into the YOLOv8n-Segment architecture. This combination led to a 5.5% increase in performance over the baseline model. Additionally, YOLOv8n-Pose was incorporated for precise keypoint-based affine correction, further enhancing the model's accuracy and efficiency. As a result, accurate and highly efficient phenotype extraction was achieved. A strong correlation ( R 2 = 0.96) was observed between model-predicted and manually measured total lengths, and body weight prediction based on image-derived traits also showed high accuracy ( R 2 = 0.854). Additionally, heritability estimates and genomic-estimated breeding values derived from deep learning-predicted weights were comparable to those obtained from manual measurements. These findings indicate that TovNet-Vision, enhanced by its uniquely designed LEC module, provides a scalable and non-invasive solution for phenotyping and holds strong potential for enabling more efficient, data-driven breeding strategies in aquaculture. The demonstrated accuracy and reliability highlight the promise of this framework as a powerful tool for molecular breeding and sustainable aquaculture.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: