Machine Learning Fisheries Monitoring

(1) Managing fisheries resources, habitats, passage structures, and restoration efforts requires monitoring and population assessments of target species. Currently humans spend long hours manually counting fish by viewing still images and video segments on computers. This process is time consuming, costly and prone to error.

(2) Our solution is to develop automated video monitoring systems for real-time 24-7 monitoring and analysis.

(3) This solution will enable fisheries managers and monitoring groups globally to easily and accurately conduct, quantify, and analyze fisheries resources in real-time, reducing cost and time, and providing more accurate information to inform resource management processes. The technology will also be used in outreach, education and training programs for stakeholders at all levels including undeserved and tribal communities.

The Problem: Fish population distributions are shifting in response to climate change (Hare et al., 2016; Nye et al., 2009), which can negatively impact fishing communities, coastal economies, and world food supplies. Estimating fisheries stocks and species distributions requires surveying and monitoring to determine species occurrence, abundance, and migratory patterns. In general, fisheries monitoring is a time consuming manual process involving large data sets prone to human error. State and federal agencies have been interested in machine learning assisted monitoring of fisheries, but development of such systems has shown limited success. Applying machine learning to fisheries management has become a priority by state and federal agencies.  

Jonathan A. Hare, et al., 2016.  A Vulnerability Assessment of Fish and Invertebrates to Climate Change on the Northeast U.S. Continental Shelf. PLOSOne  

JA Nye, et al., 2009. Changing spatial distribution of fish stocks in relation to climate and population size on the Northeast United States continental shelf. Marine Ecology Progress Series 393, 111-129

The Solution: We will develop automated video monitoring software for use in fish population monitoring and habitat assessments. Applying machine learning automated technology to fish monitoring can reduce human burden, improve efficiency and accuracy, improve assessments of fish stocks and climate impacts on fisheries and resources, and enable coastal communities to become more informed and resilient to climate impacts on coastal and marine resources. We will develop software by apply various machine learning methods in combination that will identify, quantify, and analyze fisheries populations by species, including fish condition, biological, and physical indicators, and various environmental sensors and physical metrics used in the management of fisheries and associated resources and habitats. An example of our prototype can be viewed here:

https://www.dropbox.com/sh/dh0jie3x9ucap6z/AABj1WvyLcSFKWrEBnLju4e6a?dl=0

The final software product will be open source and made freely available to fisheries managers, monitoring groups, and the general public. 

Who does the solution serve? Automated fisheries monitoring systems reduce time, effort, cost, and potential for human error. The result is improved  efficiency and accuracy of data used by resource managers in stock
assessments, resource conservationists in coastal community resource
assessments, and the general public in promoting awareness and
stewardship through education and citizen science. More accurate and
timely assessments of fisheries resources enables resource management
and industry to support sustainable resources, coastal communities, and
economies. As fish stocks shift, communities and economies on the
receding end of the species range will be negatively impacted, while
communities and economies in the expanding area will be positively
impacted.  Communities on both ends of the shifting range must be
prepared in order to adapt to change, maintain resilience, and continue
to thrive. In addition, fish stocks do not acknowledge international
boarders, which can lead to conflicts over resource allocations and
sustainable fisheries management as fisheries ranges shift across
international boarders. Machine learning enhanced monitoring and
processing enables more timely and accurate assessments of altered
fisheries distributions in response to climate change, enabling resource
managers, industry, and community planners to better prepare for
impacts to coastal communities, economies, and marine resources. We are currently working with NOAA Fisheries Service and Habitat Restoration Center, the NOAA Northeast Fisheries Science Center, Massachusetts Division of Marine Fisheries, local municipalities, local watershed and diadromous fish monitoring groups, and the Wampanoag Tribe in Massachusetts. We also work closely with the regional commercial fishing industry to assess their needs for improved monitoring technology. These are the primary end user groups that will benefit from this technology. These groups provide us with feedback on their specific challenges and needs, which guide our technology development. These groups also provide the video needed for machine learning training and system development.

We are working closely with the Mashpee Wampanoag Tribe in Massachusetts to develop the technology to help with their monitoring and assessing the effects of fish passage and restoration migratory fish populations and sustainable fisheries resources for the Tribe. Part of this collaboration includes developing citizen science monitoring programs, and education and training programs for the tribal community and tribal school groups that focus on technology, natural resources, and cultural heritage. Crowd sourcing with various tribes nationally will increase the reach of education and training programs involving fisheries resources, cultural heritage, and the technology we are developing for improving resource management
and assessment, as well as aid in product development, testing, and validation.

We are developing machine learning applications for fisheries monitoring and assessing stocks, climate impacts and resources. Currently fisheries surveys using images are manually reviewed, resulting in many hours and increased likelihood for human error.  Automation through machine learning would reduce human burden and processing time, and improve timeliness and quality of data, reduce uncertainty and error in results, and improve fisheries assessments and management. More accurate assessments support long-term sustainability of fisheries resources and improves resource allocations in support of tribal and local communities, industry, and economies. Collaborating with local tribal communities on citizen science, crowd sourcing, education, and training programs will increase awareness of natural resources and their importance with regards to tribal heritage.  The program will also provide technical training and STEM education opportunities for advancing education and career goals for members of the tribal community.

By focusing on developing sustainable fisheries resources in collaboration with tribal communities, we align with the Challenge focus areas:

• Functioning ecosystems crucial to wild food sources

• Technologies that improve assessment of fisheries resources

• Technology-based solutions that help communities restore, sustain, and benefit from resilient ecosystems. To that end, Solve

• Technology coupled with tribal community knowledge and resource improvements

We are in the process of working with federal, state, tribal, and local fisheries groups to develop machine learning software for use in fish passage and restoration monitoring, assessments, as well as education, and outreach programs.  We have developed a prototype system that is a work in progress and requires additional work to complete and distribute to end users. An example can be viewed through this link:  

https://www.dropbox.com/sh/dh0jie3x9ucap6z/AABj1WvyLcSFKWrEBnLju4e6a?dl=0

Automated monitoring systems for stock assessments and fisheries management do not currently exist. Our Solution is innovative since it will provide a technological solution that will reduce time and costs, and improve processes and quality of information needed to effectively evaluate and manage fisheries populations and the habitats and resources that support them.  The system will benefit federal and state fisheries and habitat managers and policy development processes, local monitoring groups and municipalities, and inform, educate, and train undeserved and tribal communities.

The exact number of people our solution currently and will potentially serve is difficult to say.  We collaborate with resource managers, monitoring groups, municipalities, and tribal groups, all of whom represent greater local, regional, and national, and international populations.  Therefore, we provide the following:

Currently (year one) we collaborate in Massachusetts with two federal agencies, two regional fisheries management councils, six state agencies, five water shed associations, one tribal group, and a developing collaboration in the UK.

In five years we expect to collaborate in Massachusetts with five federal agencies, three regional fisheries management councils, seven state agencies, and seven watershed management groups, and ten tribal groups (three in MA and seven nationally). In five years we also expect to expand beyond Massachusetts to all of the US coastal states, as well as with developing collaborators in the Canada and the UK.

Measurable indicators include:

• Number of collaborators actively involved
• Number of stakeholders expressing the need for this solution
• Potential population impact represented by collaborator groups
• System development rate and level of accuracy
• Level of data, process, and information improvement over current monitoring methods
• Status of end user product development
• Implementation by collaborators
• Effectiveness of improving efficiency and accuracy of information for resource assessment and sustainability management
• Product distribution rate and number of stakeholder groups using the product
• Number of countries product is in use
• Ability to acquire funding to support the effort
• Number of tribal communities involved
• Number of tribal community students engaged in education and training programs
• Number of tribal community students that pursue education and careers in technology and natural resource management as a result of participating in this program
• UN Goals
  • Ensure inclusive and equitable quality education and promote lifelong learning opportunities for all
  • End hunger, achieve food security and improved nutrition and promote sustainable agriculture
  • Promote sustained, inclusive and sustainable economic growth, full and productive employment and decent work for all
  • Build resilient infrastructure, promote inclusive and sustainable industrialization and foster innovation
  • Reduce inequality within and among countries
  • Make cities and human settlements inclusive, safe, resilient and sustainable
  • Ensure sustainable consumption and production patterns
  • Conserve and sustainably use the oceans, seas and marine resources for sustainable development
  • Take urgent action to combat climate change and its impacts*
  • Strengthen the means of implementation and revitalize the global partnership for sustainable development

1 full-time staff

1 part-time staff

2 students

Robert Vincent has thirty-years experience in fisheries and natural resource management and works closely with all collaborators, stakeholders, and resource managers to understand and develop technology solutions that address their specific needs. Robert Vincent also works closely with the Mashpee Wampanoag Tribe to develop solutions natural resource management needs and education programming.

Bran Anthony is principal research scientist in the MIT Department of Mechanical Engineering and specializes in developing machine learning applications for imaging solutions

We use the following as our guides for Diversity, Equity, and Inclusion programming:

Diversity - Sea Grant embraces individuals of all ages, races, ethnicities, national origins, gender identities, sexual orientations, disabilities, cultures, religions, citizenship types, marital statuses, education levels, job classifications, veteran status types, and income, and socioeconomic status types. Sea Grant is committed to increasing the diversity of the Sea Grant workforce and of the communities we serve.

Equity - Sea Grant provides individuals and communities a voice and opportunity in decision making. Sea Grant is committed to a policy of equal opportunity for all persons and does not discriminate. Sea Grant works to challenge and respond to bias, harassment and discrimination.

Inclusion - Sea Grant is committed to building inclusive research, extension, communication and education programs that serve people with unique backgrounds, circumstances, needs, perspectives and ways of thinking. Sea Grant cultivates a sense of belonging among staff, partners, and communities served.

At MIT Sea Grant our focus is on developing technology to solve stakeholders challenges.  SOLVE(ing) is what MIT Sea Grant is all about. Therefore, a funding opportunity that focuses on solving is an appropriate program to apply to.  The greatest barrier for our program to accomplish its goal is funding, which is why this has been a primarily a side project for several years.  In other words, as small pools of funding have emerged over the past several years we focus our efforts, but what we require is a substantial funding commitment to support a focused effort on development, testing, and finalization, production, and distribution of the product.

We are open to partnering with others that can provide expertise in program impacts and evaluations, as those who have expertise in applying machine learning to imaging challenges.

We would like to partner with organizations, MIT Faculty, or SOLVE member with expertise in machine learning for imaging and who have a commitment work with us until the product is completed.  Partners would also have to agree that the product, software, etc. will be open source and made freely available for all end users.  We are creating a technological solution to give to the global community.

We would use the funding from the Innovation for Women Prize to support one woman graduate student to specialize in machine learning applications to imaging and work with us on developing our machine learning for fisheries monitoring product.

Our team is focused on developing technology to improve monitoring, assessment, and management of fisheries resources.  In doing so, our technology will support more informed fisheries management leading to more effective sustainability policies and management, contributing to the reduction and end of global overfishing. The funding from the Minderoo Prize will be used to develop and distribute the technology and train end users for global implementation.

Our solution is based on using AI applications for development of improved fisheries monitoring and management technology. The resulting technology will improve fisheries assessments and management, improving sustainability
of fisheries resources and food supplies for the global humanity. We will use funding from the AI for Humanity Prize to support development, testing, distribution, and training of the product globally.

Our Solution focuses on combining several machine learning applications to the development of innovative fisheries monitoring and assessment technology.  Pressure on fisheries resources is an issue both on the global and local community level, and development of technological solutions to improve our ability to effectively monitor, assess, and sustainably manage fisheries resources is of importance to global food supplies.  Our program collaborates with local communities, as well as local tribal communities, to engage students and adult learners in STEM education and training with an emphasis on preparing students for careers in technology, and training the next generation of scientists and engineers. We will use funding from the GSR Prize to support development of the technology and product, and for developing education programming with tribal communities, undeserved communities, and undergraduate student research positions in support of our work.