By leveraging the same technology that is helping power self-driving cars, drones and fully autonomous vehicles, the Synthetik team is working with NOAA to develop a system to mitigate marine entanglement events at aquaculture facilities and other high-risk man-made marine obstacles.
This system, DeepSeaVision-AI (DSV-AI), uses a suite of modern sensors to observe areas in and around a marine aquaculture facility. This sensor data is processed using advanced computer vision and machine learning, and when an endangered marine animal is detected, a warning system is triggered to help guide the animal away from the area, and prevent a potentially deadly entanglement event.
DeepSeaVision-AI: Scalable Deep Learning to Support Animal Ecology
1. Scalable Deep Learning to Support Animal Ecology
DeepSeaVision-AI:
Photo by aprilandrandy / CC BY
2. Entanglement of endangered marine animals with offshore aquaculture equipment is a serious problem -
and it’s only getting worse as we expand into deeper offshore environments to keep up with the global
demand for seafood. Synthetik is using AI to save over 115 marine species from entanglement, whilst
assisting the sustainable development of aquacultures in the future.
Photo by NOAA PIFSC / CC BY
3. Problem
• Population growth coupled with the over-exploitation of wild fish stocks and declining water
quality, means that the global demand for seafood now exceeds the production capacity of our
natural ocean ecosystems. Consequently, there has been explosive growth in offshore aquaculture
to attempt to meet consumer demand, with improved technology allowing marine aquaculture to
expand into deeper, previously inaccessible, offshore environments.
• These new facilities pose a serious entanglement risk, which is particularly acute for many
endangered marine species. In the United States alone, at least 115 species of marine mammals,
sea turtles, sea birds, fish, and invertebrates are affected, and this problem will only get worse as
we continue to develop and expand marine aquaculture.
• Whilst determining the exact frequency of marine entanglement is difficult due to a lack of data, a
review of global turtle bycatch reporting efforts between 1999 and 2008 by Wallace et al. found
approximately 85,000 turtle bycatch incidents reported worldwide, with estimates suggesting that
due to underreporting, the actual number could have been greater than that by over 2 orders of
magnitude. It is clear that our fishing efforts pose a serious risk to future of many of our most-
endangered marine species.
4. What we are doing
• We are using AI to help save endangered marine animals.
• We are creating new systems to prevent entanglement and limit
the impact of our fishing practices on the marine environment.
• We are using technology to support the future sustainable
development of aquacultures to help meet the global requirement
for seafood without further endangering at-risk marine species.
• We’re developing industrial-scale solutions to meet industrial-scale
challenges.
5. How we are doing it
• By leveraging the same technology that is helping power self-driving cars,
drones and fully autonomous vehicles, the Synthetik team is working with
NOAA to develop a system to mitigate marine entanglement events at
aquaculture facilities and other high-risk man-made marine obstacles.
• This system, DeepSeaVision-AI (DSV-AI), uses a suite of modern sensors to
observe areas in and around a marine aquaculture facility. This sensor data is
processed using advanced computer vision and machine learning, and when
an endangered marine animal is detected, a warning system is triggered to
help guide the animal away from the area, and prevent a potentially deadly
entanglement event.
7. Who we are doing it with
In addition to our work with NOAA in developing the DeepSeaVision-AI (DSV-AI) system:
• We are interacting with global experts and institutions to raise awareness, implement best
practice, and produce an optimized and universally compatible solution.
• We are speaking with search engines and social media platforms to support our innovative
approach to crowd-sourced data processing.
• We are discussing commercial backing, affiliation, funding, and hardware donations with
global technology companies.
8. Challenges
• The marine environment itself poses significant challenges, requiring special consideration
to ensure the DSV-AI solution can survive and operate under adverse environmental
conditions.
• Training DSV-AI’s deep learning-based detection, classification, and tracking models require
a significant amount of both raw training data and manpower to process and annotate it -
this means finding creative ways to effectively leverage crowd-sourced data processing.
• As marine aquaculture is moving further and further offshore, communication and data-
transfer become increasingly restricted, meaning that sensor data needs to be processed
in-place using embedded hardware, and robust fully-autonomous decisions must be taken
without human interaction.
9. Summary
• Despite these challenges, and with the advent of improved artificial intelligence,
particularly with respect to deep machine learning and computer vision - coupled with
inexpensive yet powerful embedded computing hardware and sensors, the Synthetik team
are confident that intelligent, scalable, and effective solutions to these difficult problems
can be found.
• Our mission is to use technology to develop scalable solutions to help humanity, and our
environment, thrive now and in the next century.
• We believe that with smart, AI-driven, and truly scalable solutions, we can help bring back
endangered marine species and eliminate unnecessary, tragic, and entirely preventable
marine entanglement events for good.