用AI自动分析南美白对虾形态，提升育种效率。
IMASHRIMP: Automatic White Shrimp (Penaeus vannamei) Biometrical Analysis from Laboratory Images Using Computer Vision and Deep Learning
- 基于改进ResNet-50和VitPose，实现视角分类与骨骼点定位。
- 姿态估计mAP达97.94%，像素转厘米误差仅0.07cm。
- 双认证系统降低人工错误，适合水产育种与自动化研究。
本文提出IMASHRIMP系统，用于从RGBD图像中自动分析南美白对虾（Penaeus vannamei）的形态特征，以优化水产养殖中的遗传选育。针对虾类形态分析的特殊挑战，对现有深度学习与计算机视觉技术进行适配。系统包含两个基于改进ResNet-50的判别模块，分别用于图像视角分类和额角完整性判断；引入“人机双因子认证”机制，将视角分类的人工错误率从0.97%降至0%，额角检测错误率从12.46%降至3.64%。同时，基于VitPose改进的姿势估计模块可预测23个关键点，分别处理侧视与背视图像。采用支持向量机（SVM）模型构建形态回归模块，实现像素值到厘米单位的转换，误差为0.07（±0.1）cm。实验表明，系统显著降低人为误差，姿态估计的平均精度（mAP）达97.94%。该系统具备自动化、高精度特点，可加速虾类形态分析，助力可持续水产养殖发展。代码开源：https://github.com/AbiamRemacheGonzalez/ImaShrimp-public
原文摘要 · Abstract (English)
This paper introduces IMASHRIMP, an adapted system for the automated morphological analysis of white shrimp (Penaeus vannamei}, aimed at optimizing genetic selection tasks in aquaculture. Existing deep learning and computer vision techniques were modified to address the specific challenges of shrimp morphology analysis from RGBD images. IMASHRIMP incorporates two discrimination modules, based on a modified ResNet-50 architecture, to classify images by the point of view and determine rostrum integrity. It is proposed a "two-factor authentication (human and IA)" system, it reduces human error in view classification from 0.97% to 0% and in rostrum detection from 12.46% to 3.64%. Additionally, a pose estimation module was adapted from VitPose to predict 23 key points on the shrimp's skeleton, with separate networks for lateral and dorsal views. A morphological regression module, using a Support Vector Machine (SVM) model, was integrated to convert pixel measurements to centimeter units. Experimental results show that the system effectively reduces human error, achieving a mean average precision (mAP) of 97.94% for pose estimation and a pixel-to-centimeter conversion error of 0.07 (+/- 0.1) cm. IMASHRIMP demonstrates the potential to automate and accelerate shrimp morphological analysis, enhancing the efficiency of genetic selection and contributing to more sustainable aquaculture practices.The code are available at https://github.com/AbiamRemacheGonzalez/ImaShrimp-public
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。