How Artificial Intelligence Can Improve Rotifer Production for Aquaculture

By Jia Geng

For most marine fish species, fish larvae can only feed on small particles that fit their mouth gap. Rotifers, a microscopic plankton species, are a crucial starting diet for marine fish because of their suitable size.

In commercial-scale fish farms, however, each production cycle could require billions of rotifers as live feed. So a sufficient and stable supply of high-quality rotifers is one of the key elements to enhancing fish production.

Challenges in Rotifer Cultivation

Rotifers are usually cultivated in high-density batch or semi-batch cultures using formulated enrichment media as their food source and chemicals to control ammonia and maintain proper pH levels.

Commercial-scale rotifer cultivation requires experienced culturists to closely observe the rotifer culture and collect information such as rotifer density, fertilization rate and swimming speed. The amounts of input diet and chemicals need to be carefully calibrated based on the collected information.

Observation and data collection are labor-intensive and inefficient, however, with rotifer culturists spending hours each day sampling, observing and counting rotifers under microscopes to obtain the data needed to manage the rotifer culture accordingly.

AI’s Role in Rotifer Cultivation

In recent years, the advancement of machine learning algorithms has greatly enhanced the capabilities of machine vision, especially in applications such as object detection, localization, classification and tracking. Case in point: Innovasea’s feed optimization and biomass estimation capabilities.  

An artificial intelligence-driven machine vision system designed for rotifer cultivation could help culturists efficiently observe, interpret and evaluate the cultivation process by analyzing the microscopic images of rotifer samples. For instance, AI could be used to:

  • Estimate Rotifer Density via Object Detection – Rotifer density is the most important metric in rotifer culture management because rotifer culturists decide the feed and chemical inputs based on the estimated population. Object detection algorithms can be trained to recognize rotifers in microscopic images so the system can estimate rotifer density by counting the detected rotifers in fixed volume samples.
  • Estimate Rotifer Fertilization Rate via Object Classification – The rotifer fertilization rate can help culturists evaluate the state of the cultivation, assess the population growth and make strategic plans on larvae feeding. Fertilized rotifers can be identified by the eggs attached to their bodies. As a common practice, rotifer culturists need to count the fertilized rotifers and unfertilized rotifers in the samples to estimate the fertilization rate. With object classification algorithms, machine vision could further distinguish the detected rotifers into fertilized rotifers and unfertilized rotifers so that the system could estimate the fertilization rate like rotifer culturists do.
  • Estimate Rotifer Swimming Speed via Object Tracking – Swimming speed is an effective indicator of the healthiness of rotifers. Rotifers in sub-optimal conditions usually exhibit slower swimming speed. Rotifer culturists can tell whether rotifers are swimming normally based on their experience. However, the assessment is subjective and could lead to inconsistent management. Machine vision with object tracking algorithms could analyze image sequences to track the swimming rotifers. The rotifer trajectories could be used as objective measurements to evaluate rotifer swimming speed.

Single-Shot Measurement vs. Continuous Measurement

Ideally, continuous measurement and monitoring could provide more information about the cultivation and reduce the production risk. However, commercial hatcheries usually cannot afford continuous monitoring. As a result, the observation and measurements are typically conducted in a single-shot manner. For example, rotifer culturists count rotifers and take samples only once in the morning – and then make their operational decisions based on that snapshot of information so they can get back to performing other tasks in the hatchery.

AI-driven machine vision would enable hatcheries to continuously monitor and measure the key indexes of rotifer cultivation so that the rotifer culture could be managed in a more precise way. The continuous data collection could also expand the rotifer culturists’ knowledge base and help improve the rotifer cultivation protocols.

Just like advances in AI are helping improve other aspects of aquaculture, such as feeding and fish monitoring, the technology could improve the efficiency and quality of rotifer culture, decrease the need for human labor, and reduce the cost of the commercial-scale rotifer production.

About the Author

Jia Geng is a Ph.D. candidate in the department of Marine Ecosystems and Society at the University of Miami. He received his Bachelor of Science in aquaculture from Ocean University of China and a graduate certificate in data science foundations from North Carolina State University. His research focuses on developing machine learning-based computer vision systems for rotifer culture.

The viewpoints and opinions expressed in this article are those of the author and do not necessarily reflect those of Innovasea or its employees.


You are using an outdated browser. Things may not appear as intended. We recommend updating your browser to the latest version.