Introduction
The implementation of genetic breeding programmes (GBPs) in the Penaeus vannamei production sector has reduced generation intervals and demonstrated significant genetic gains for traits related to growth and survival (Castillo-Juárez et al., 2015; Martínez Soler et al., 2024). External morphometric traits—reflecting the species’ health status under specific culture conditions (Singh et al., 2017)—have emerged as valuable alternative to traditional body weight measurements for improving shrimp performance. These traits exhibit high heritability (h²), strong positive genetic correlations (rg) with both weight and product quality traits and offer high measurement precision (Shin et al., 2023; Martínez Soler et al., 2024). Moreover, they can be measured non-invasively through image analysis technologies. The objective of the present study is to integrate the IMASHRIMP system—which combines deep learning and computer vision—as an advanced tool to enhance genetic selection of P. vannamei within the PMG-BIOGEMAR© GBP at commercial scale.
Materials and methods
588 shrimp from the sixth generation (G6) of the PMG-BIOGEMAR© GBP—genetically selected for growth traits and reared under commercial conditions—were photographed at harvest size using an Intel® RealSense™ D435 depth camera. Each shrimp was imaged from both lateral (right and left) and dorsal views, with four different angles captured per view, resulting in twelve images per individual. The initial module of the IMASHRIMP system uses pose estimation algorithms to automatically identify the shrimp’s orientation (lateral or dorsal view; right or left side) and detect the presence or absence of the rostrum. Based on this information, the system applies one of two customized skeletal key point models—depending on the detected view and rostrum presence—to extract morphometric traits in pixels. A second module then converts the pixel-based measurements into centimetres using a Support Vector Machine (SVM) regression model. The resulting traits are classified as non-invasive morphometric traits (mNiT). To validate the IMASHRIMP system, the 588 shrimp were manually measured for the same 22 morphometric traits—designated as invasive morphometric traits (miT)—as well as for body weight. These traits, related to length, width, and height, were previously defined and validated by Shin et al. (2023) and Martínez Soler et al. (2024). All shrimp were genotyped using the AQUAarray LDTM SNP-chip at the Center for Aquaculture Technologies (California, USA). Additive genetic components for all morphometric traits were estimated using the Restricted Maximum Likelihood (REML) method. Best Linear Unbiased Prediction (BLUP) evaluation was performed using VCE (version 6.0) and supporting software.
Results
At 103 days after hatching (DAH), the weight and total length (TL) of shrimp were 23.74 g and 14.35 cm, respectively, following five generations of genetic selection. Shrimp analysed were assigned to 135 full-sib and 25 half-sib families. h² for weight was 0.37. For miT, h² values ranged from 0.05 to 0.40 for length traits, while for mNiT, they ranged from 0.14 to 0.36. The lowest h² was observed for first segment length (SL1), and the highest for cephalothorax length (CL) and sixth segment length (SL6) in both miT and mNiT. For height traits, h² ranged from 0.13 to 0.30 for miT and from 0.21 to 0.41 for mNiT. When shrimp were measured using IMASHRIMP system from a lateral view, all morphometric traits—except CL—showed increased h². The average increase for length traits was 90%, while the average increase for height traits was 37.40%. When mNiT measurements were taken from a superior view, h² values for length ranged from 0.13 to 0.45, with the highest value found for abdominal length (AL). In this case, h² increased by an average of 95% compared to miT. Regarding width traits measured from a superior view, h² ranged from 0.08 to 0.40 for miT and from 0.10 to 0.33 for mNiT. rg between miT and mNiT ranged from 0.81 to 1.00 for lengths (lateral view in mNiT), from 0.53 to 1.00 for heights, from 0.72 to 0.98 for lengths (superior view in mNiT), and from 0.98 to 1.00 for widths. rg between mNiT and weight ranged from 0.82 to 1.00.
Discussion
Shrimp from G6 exhibited 30% higher body weight 27 days earlier in commercial estuaries compared to second generation (G2) (Shin et al., 2023), demonstrating that genetic selection is significantly accelerating growth under farming conditions. However, increases in length were less pronounced (8%) compared to width (15%) and height (16%). The inclusion of morphometric traits as selection criteria is particularly relevant for shrimp markets in the USA and Europe, where whole animals are marketed and consumers place high value on exoskeletal integrity. But comprehensive morphometric assessments in animal production are typically labour-intensive, time-consuming, and costly. Recent advances in deep learning and computer vision (Chai et al., 2021) offer new opportunities for aquaculture by enabling high-throughput, cost-effective measurement of traits that were previously difficult to assess. The IMASHRIMP system represents a robust technological tool for integration into the sampling workflows of shrimp GBPs. To the best of our knowledge, this is the first application of pose estimation in the shrimp aquaculture sector. IMASHRIMP significantly reduces sampling time, minimizes measurement error, and increases the h² of key morphological traits in P. vannamei. Successfully integrated into the PMG-BIOGEMAR© GBP, this system enhances selection accuracy and supports long-term genetic improvement.
Acknowledgments
This study has been supported by BIOGEMAR S.A. (Reference: C2021_72).
References
Chai, J., Zeng, H., Li, A., Ngai, E.W.T., Deep learning in computer vision: A critical review of emerging techniques and application scenarios. Machine Learning with Applications. 100134.
Shin, H.S., Montachana Chimborazo, M.E., Escobar Rivas, J.M., Lorenzo-Felipe, A., Martínez Soler M., Zamorano, M.J., Fernández, J., Ramírez Artiles, J., Peñate Sánchez, A., Lorenzo Navarro, J., Intriago Díaz, W., Torres, R., Reyes Abad, E., Afonso, J., 2023. Genetic parameters for growth and morphometric traits of the Pacific white shrimp Penaeus vannamei from a selective breeding programme in the industrial sector of Ecuador. Aquaculture Reports 31, 101649.
Singh, A., Datta, S.N., Ansal, M.D., 2017. Biometric characteristics of Pacific white shrimp, Litopenaeus vannamei (Boone, 1931) cultured in the salt affected area of district Fazilka (Punjab) India. Indian J. Ecol. 44 (3), 628–631
Martínez Soler, M., Shin, H.H., Lorenzo-Felipe, A., Zamorano, M.J., Ginés Ruiz, R., Pachón Mesa, L.C., González, D., Fernández, J., Ramírez Artiles J.S., Peñate Sánchez, A., Lorenzo Navarro, J., Torres, R., Reyes Abad, E., Afonso, J.M., 2024. Genetic parameters of meat quality, external morphology, and growth traits in Pacific White Shrimp (Penaeus vannamei) from an Ecuadorian population. Aquaculture 593, 741228.