MortCam: a tool for fish mortality monitoring in RAS

Photo of author

By Milthon Lujan

The MortCam deployed in the (a) growout tank to acquire the mortality data above the drain plate in (b) ambient light (AL) and (c) supplemented light (SL) conditions. Source: Ranjan et al., (2023)
The MortCam deployed in the (a) growout tank to acquire the mortality data above the drain plate in (b) ambient light (AL) and (c) supplemented light (SL) conditions. Source: Ranjan et al., (2023)

One of the most crucial indicators in the aquaculture industry is fish mortality. Unusual mortality patterns can be associated with abiotic or biotic factors in recirculating aquaculture systems (RAS).

In RAS, monitoring tank drains is critical to maintaining system operations. The accumulation of dead fish at the tank bottom can lead to drain clogging and subsequent complications.

A team of researchers from The Conservation Fund Freshwater Institute developed MortCam, a tool for automated mortality monitoring in RAS.

The challenge of monitoring mortality

For aquaculture farm managers, monitoring mortality is essential for making informed decisions about RAS management and addressing underlying causes to prevent mass mortality events.

Traditional methods often rely on periodic observations by human operators and underwater cameras. However, this approach has limitations as it does not provide continuous real-time monitoring.

MortCam: a revolutionary solution

MortCam is an innovative solution that combines artificial intelligence (AI) and the Internet of Things (IoT) to offer 24/7 real-time mortality monitoring. This tool consists of an integrated imaging sensor with edge computing designed specifically for underwater applications.

The MortCam tool was deployed in a 150 m³ circular RAS tank for salmon farming (Salmo salar) at 0.6 meters above the bottom drain plate to acquire image data under both ambient and supplemental light conditions. Images were captured every fifteen minutes over a period of 90 days.

See also  Optimizing Microalgae Cultivation for Biodiesel Production

Acquired images were labeled as “alive” or “dead” and split into training (70%), validation (20%), and test (10%) datasets to train a custom YOLOv7-based mortality detection model.

Promising results

The optimized model achieved a mean average precision (mAP) of 93.4% and an F1 score of 0.89, demonstrating its ability to accurately and reliably detect mortality.

Additionally, the model showed robustness in varying imaging conditions, which is essential for application in diverse environments.

Practical application in mortality monitoring

“Data analyses indicate that, while deploying the mortality models on corresponding test images, predicted mortalities by the model and actual mortalities were significantly correlated,” reported the researchers.

The model was implemented in MortCam to achieve round-the-clock autonomous mortality monitoring. The system reliably generated email and text alerts to notify production staff of unusual mortality events.

“In general, the developed models performed well in terms of mortality tracking when tested on images acquired under similar conditions. However, the model’s performance substantially degraded when mortality models were cross-validated,” the researchers emphasized.


MortCam represents a significant advancement in aquaculture by providing a real-time and precise mortality monitoring tool. This technology enables farm managers to take proactive measures to address mortality issues and improve fish welfare. Furthermore, MortCam’s adaptability to various imaging conditions ensures its utility in a variety of aquaculture environments.

Key findings of the study include:

  • Mortality models trained with images captured under specific lighting conditions (i.e., AL or SL) performed well in terms of mAP, F1 score, and mortality count when validated under similar conditions but had reduced performance when cross-validated (i.e., AL model validated on SL images and vice versa).
  • A mixed mortality model trained with diverse image datasets (i.e., AL and SL images) performed better than the respective AL and SL models. The mixed model’s performance remained robust even in changing lighting conditions.
  • MortCam facilitated 24/7 autonomous mortality monitoring and reliably alerted the crew to unusual mortality events.
See also  Benefits of aquaculture nutritional research to meet the challenges of the 21st century

With MortCam, the aquaculture industry is taking a significant step towards more efficient and ethical production.

The study was funded by the USDA Agricultural Research Service.

Rakesh Ranjan
The Conservation Fund Freshwater Institute
Shepherdstown, WV, 25443, USA
Email: rranjan@conservationfund.org

Reference (open access)
Rakesh Ranjan, Kata Sharrer, Scott Tsukuda, Christopher Good. 2023. MortCam: An Artificial Intelligence-aided fish mortality detection and alert system for recirculating aquaculture, Aquacultural Engineering, Volume 102, 2023, 102341, ISSN 0144-8609, https://doi.org/10.1016/j.aquaeng.2023.102341.

Leave a Comment