SnappArt 360 E-SENSE: AI Model for Automated Egg Stage Detection to Support Modern Hatchery Management

Accurate evaluation of spawned egg numbers and embryonic development is critical for hatchery management and larval survival. Traditional manual stereomicroscope assessments remain subjective, labor-intensive, and prone to variability. To address these operational challenges, the SnappArt 360 E-Sense automated imaging system was developed. Combining high-resolution imaging on an Nvidia Jetson edge-computing platform with a YOLO-based object detection model, the system identifies and classifies multiple stages: viable fertilized eggs, developing embryos, degenerated eggs, larvae, membranes, and oocyte maturation stages (A-D) for hormonal induction planning. Model training utilized expert-annotated images via the CVAT platform across marine fish species including Sparus aurata, Dicentrarchus labrax, Lates calcarifer, Solea senegalensis, Scophthalmus maximus, and Gadus morhua. Field trials demonstrated 100% correct egg detection up to 200 items per well, achieving an overall counting accuracy of 99% while cutting sample processing time by approximately 50% compared to manual methods. This AI-driven tool standardizes egg quality assessments, enhances traceability from fertilization to hatch, and optimizes broodstock management decisions.