Team Name: OnDeck Fisheries AI
Concept: For fisheries, OnDeck is a monitoring tool that satisfies regulatory needs using AI instead of humans to automatically quantify catch; revolutionizing marine conservation while cutting costs by a factor of 10.
Challenge: Fisheries support the livelihoods of 10% of the world’s population, generating as much as $400 billion USD in revenue annually. Fish provide more than 2.9 billion people with 20 percent of their animal protein needs, and are an irreplaceable source of nutrients to many coastal communities. Despite the importance of the world’s fish stocks, two thirds of fish stocks are currently in decline, threatening the global economy and billions of livelihoods. Quantifying the number of fish taken from the ocean is critical to fisheries management, however it is not currently possible to measure this at scale.
Currently, the industry uses human observers on board a small percentage of fishing vessels to monitor what goes on at sea. These observers operate in challenging and hostile conditions where they are frequently threatened by the crew, and research has found that their collected observations can contain significant misreporting. These human observers cost between $500-$800 per day, paid for by the captain of the ship.
Recently, cameras have been installed on some boats to replace human observers. These cameras (known as “electronic monitoring”) film for entire fishing trips, which vary in length from 1-8 weeks. Once the vessels return to shore, reviewers watch the footage and manually identify the amount and types of sea life brought on board. This is extremely expensive and slow, creating a significant delay in reporting.
Solution: The majority of tasks performed by human observers or reviewers can easily be automated. Leveraging modern software technology, specifically machine learning and computer vision, we will automate the counting of fish in the videos recorded by existing electronic monitoring hardware. Our idea is to develop a software product which can automatically analyze video footage of fishing trips, measuring the number and types of fish caught as well as identifying bycatch when it occurs.
Our solution has many advantages over the manual analysis being used currently. Automated monitoring is nearly instantaneous, much faster than manual review which can take hours to days. By virtue of its speed and the inherently low marginal cost per product of software, automated monitoring will be far less expensive than manual review. Likewise, the increased speed and decreased cost will allow for comprehensive monitoring of a far larger portion of global fishing activity, which will be integral to improving the sustainability of the fishing sector in the coming decades. Finally, we note that human observers are subject to human error, which can lead to inaccuracy and bias. We believe that automated monitoring software can achieve >95% accuracy with minimal bias. This level of accuracy is well within reach of existing computer vision algorithms.
Market: There are currently over 1,400 fishing vessels with electronic monitoring (EM) systems on board today, and this number is expected to grow to 25,000-50,000 by 2028. With this explosion of EM, automation is critical to enable the review of this huge volume of data. Globally, approximately $1.5 billion USD is spent on fisheries observations annually, with at least $400 million USD spent on manual review.
Our first customers would be the established observing companies in North America. There are approximately a dozen observing providers in North America, the biggest of which are all trying to make the shift to EM. Our software would accelerate this shift and revolutionize their workflow. We have a budding partnership with the dominant observing provider in Canada to deploy our software on EM systems across Canada, the US, Australia, and New Zealand.
After 2-5 years of working with established observing providers and proving the efficacy and accuracy of our automated review software, we could become our own certified fisheries observing company. This would enable us to sell our monitoring solutions directly to fishers, facilitating vertical integration.
Competition: There is no software company developing the solutions needed for the fishery observing industry. Academic researchers have developed and proven the capabilities of computer vision technology for our use cases, but there has been no attempt to bring these solutions to market.
The four biggest observing companies in North America are making slow attempts to add some automation into their workflow, but they do not have the in-house technical capabilities to develop true computer vision or AI. These four competitors are also our main customers. We have already developed a working relationship with several industry leaders, and they are eager to purchase automation software from OnDeck Fisheries AI.
Value Creation: The potential for value creation is twofold. First and foremost, by improving global fisheries monitoring capabilities, our idea has the potential to revolutionize the management of fish stocks all over the world. Better measurement of the number of fish being taken from the ocean will help to guide policies to increase the sustainability of global fish stocks. Likewise, automated bycatch detection will aid in identifying fishing practices with significant collateral damage and thus reduce the damaging effect of fisheries bycatch on endangered species.
Secondly, this project has the potential to be highly profitable. Current fisheries monitoring requires paying by the hour for human observers to either observe on fishing vessels in real time, or manually view hundreds of hours of video footage for each fishing trip. With this current business model, the marginal cost of observing additional fishing trips is high. Our proposed automated electronic monitoring software, however, would reduce marginal costs of increased monitoring to essentially zero. Thus, we will be able to undercut the pricing of traditional monitoring services, allowing us to quickly scale up to reach a large section of the market.
Team: Our team members’ respective credentials showcase our collective business acumen (A. Dungate), scientific research expertise (M. Leighton), and software development experience (S. Dyanatkar).
Specific skills and experience that will be integral to the successful development and implementation of our vision include, but are not limited to:
– Product management experience (A. Dungate)
– Contacts and network within the fisheries sector (A. Dungate)
– Scientific research expertise (M. Leighton)
– Proven history of securing federal grant money (M. Leighton)
– Experience with machine learning and computer vision (M. Leighton)
– Software development experience at a major global software company (S. Dyanatkar)
– Previous experience in the development of innovative new technologies synthesizing software and hardware (S. Dyanatkar)
Comments or questions about this article? Email us.