Novel Algorithm to Detect, Classify, and Count Mussel Larvae in Seawater Samples Using Computer Vision

Abstract

The European Union’s mussel production industry is dependent on obtaining mussel larvae as seed for cultivation, a process traditionally monitored through labor-intensive manual sampling and microscopic counting prone to human error and time-consuming procedures. To address these challenges, our research presents a computer vision-based methodology for accurately identifying, classifying, and quantifying mussel larvae individuals across various developmental stages from microscopic images of water samples. Utilizing a neural network architecture derived from the YOLO method, our approach integrates convolutional, pooling, and fully connected layers to automate detection, classification, and accounting tasks. Through training with manually labeled samples and employing data augmentation techniques, we established a robust framework capable of processing diverse larval specimens effectively. Our research not only streamlines mussel larvae monitoring processes but also underscores the potential of computer vision techniques to enhance efficiency and accuracy in aquaculture industries.