New Tool Counts Oyster Seed Faster

Researchers at the Louisiana Sea Grant Oyster Research Lab and Louisiana State University’s Math Consultation Clinic (MC²) have developed an automated, image-based tool that counts juvenile oyster seed faster and more accurately than traditional methods.

The aim is to replace time-consuming hand-counts with a robust image-segmentation system that identifies and sizes oyster seed from photographs. The team trained an artificial intelligence deep-learning model (StarDist) on thousands of annotated images across two seed sizes, 2-4 millimeter and 4-6 millimeter. Accuracy has reached 90 to 94 percent.

“Manual counting has always been essential but incredibly tedious and prone to error,” said Elizabeth Robinson, director of the oyster lab. “This tool gives us consistent, reliable numbers in a fraction of the time.”

To make the technology accessible to hatchery workers and oyster farmers, the team created a simple interface that returns counts within seconds. Tests show automated counts take one minute, compared to three minutes for manual counting, with far less variability.

To support field use, MC² built a portable Raspberry Pi–based computer with a screen and protective 3D-printed housing. Color-contrasting 3D-printed phone platforms help standardize images, though farmers can use the software without any custom hardware. All components and code are open-source and accessible through the Aquaculture Information Exchange.

“Our goal was to make this completely farmer-friendly: open, modifiable and practical for any nursery operator,” said Sarah Bodenstein, former postdoctoral researcher at the lab and current assistant extension educator and regional aquaculture liaison at Connecticut Sea Grant. “If growers want to adapt it, they can. Nothing is locked behind proprietary systems.”

Currently counting seed from 2–6 millimeter, the researchers are refining the software to handle even smaller seed. Users can also retrain the system by manually highlighting new shapes or sizes, allowing customized model updates without creating an entirely new program.

Louisiana, one of the nation’s leading oyster-producing states, stands to benefit as hatcheries seek more efficient ways to manage inventory amid stressors such as storms and shifting salinity. Researchers are also exploring applications beyond oysters, including automated algal cell counting.

“This technology supports real-world production decisions, from estimating inventory to planning nursery schedules,” Robinson added. “It’s a step forward for oyster aquaculture.”