Page 40: of Marine Technology Magazine (November 2025)

SYRENE UNDERWATER AI CAMERA © Adobe Stock/wkproduction

Marine ecosystems can be highly susceptible to invasive species.

SYRENE:

AN UNDERWATER EMBEDDED

ARTIFICIAL INTELLIGENCE CAMERA

FOR INVASIVE FAUNA MONITORING

By Laurent Gautier, Stephane Barbot, and Damien Le Vourc’h nvasive species have a major impact on marine ecosystems tive, real-time monitoring would enable rapid detection of their by disrupting the natural balance and biodiversity. Intro- appearance and assessment of their impact. A solution based on duced mainly through human activities such as maritime embedded AI, capable of continuously analyzing images over a transport, ? shing and aquaculture, these invasive species long period of time, offers a promising alternative to post-process

Ican supplant endemic species and cause changes in the detections running on data centers, without energy constraints.

functioning of ecosystems. These changes can have signi? cant The team focused on Pterois volitans/Miles (lion? sh), a spe- negative consequences, both ecological and socio-economic, cies widespread across the globe. Accidentally introduced in including biodiversity loss, leading to a decline in species the early 1990s to the Caribbean Sea and western Atlantic, it stocks. In the Mediterranean Sea, one of the most affected re- has invaded this area where it causes serious damage to marine gions, the introduction of invasive species has already caused ecosystems. For example, the biomass of small ? sh has declined profound changes in marine habitats and food webs. This am- by 65% in the Bahamas reefs in just two years due to its intense pli? es the vulnerability of ecosystems to global changes. predation. It disrupts biodiversity and local food chains. [1]

The SYRENE project and the need for an SYRENE V1, a lion? sh detector

The SYRENE prototype was initially developed in 2020 as invasive fauna active monitoring

Started in 2020, the SYRENE project (SYstème in situ de a demonstrator to illustrate the capabilities of embedded deep

REconnaissance en temps réel de faune invasive sous-mariNE) learning image processing. The inner hardware architecture is aims to use embedded arti? cial intelligence to help local authori- simple and low cost, with an electronic board (Raspberry Pi ties detect and quantify invasive underwater species that threaten 4B) allowing connection to an Intel Neural Stick V2 VPU (Vi- the balance of local marine ecosystems. Arti? cial intelligence sion Processing Unit), which accelerates the processing of im- makes it possible to detect complex shapes that were impossible ages or videos (640*480 pixels) from a Raspberry Pi OV5647 to distinguish with any degree of accuracy just a few years ago. 5 Mpixels camera module.

Some invasive species, such as lion? sh, are active at night or As image processing requires signi? cant computing power, at dusk, making them dif? cult to count or detect by divers. Ac- the challenge was to achieve minimum performance on con- 40 November/December 2025

MTR #8 (34-47).indd 40 MTR #8 (34-47).indd 40 11/20/2025 5:49:47 PM11/20/2025 5:49:47 PM