SalmonVision Collaborative

Wild salmon are foundational for social-ecological systems around the North Pacific Rim. Having supported vibrant Indigenous fisheries for over 10,000 years, they continue to sustain economic opportunities, underpin cultures, and fuel ecosystems. Yet in an era of accelerating climate change, salmon ecosystems are experiencing rapid changes and more frequent disturbances, posing unprecedented challenges to their resilience. Across much of their range, wild salmon have experienced a widespread decline in abundance and productivity, becoming increasingly unpredictable in their returns. Yet wild Pacific salmon have persisted through multiple periods of glaciation, are biologically equipped for rapid evolution and recovery, and remain capable of supporting livelihoods and food security around much of their native range. 

The challenge of maintaining opportunities for salmon fishing has been compounded by the predominance of mixed-stock fisheries that harvest indiscriminately from numerous co-migrating populations and is further heightened by major data gaps in monitoring including the traditionally high costs of producing in-season information to support timely management decisions. These challenges are particularly acute for First Nations and Tribes who depend on salmon for livelihoods, culture, and food security, but often fish in rivers and nearshore areas after marine fisheries have already harvested returning salmon. Indigenous Nations across the NE Pacific have faced curtailed fishing seasons, impacting local economies and knowledge transmission, and increasing food insecurity. 

Along the north and central coast of BC, monitoring, assessment, and stewardship are limited by high staffing and operational costs for working in remote coastal and interior watersheds. Currently, many First Nations monitor spawning salmon abundance with in-river video or sonar cameras, however reviewing video or sonar files from these programs to produce annual estimates of salmon returns requires thousands of hours of staff time and data are typically unavailable until long after the fishing season is complete. These constraints have limited the ability of fishery managers to make informed and sustainable in-season management decisions and undermined First Nation access to fishery opportunities. In BC, emerging co-governance agreements mean that Indigenous fisheries managers contribute to setting harvest targets for commercial and recreational fisheries, and increasingly make the call on emergency closures and the authorization of fishery openings. 

AI technologies such as computer vision are proliferating rapidly, showing immense promise in commercial and public applications, as well as a number of well-documented risks. One critical challenge for the ethical development of AI is that the needs and vision of local and Indigenous communities are rarely at the forefront of technology design and development. Co-developing AI-powered tools for analysis and interpretation of real-time population monitoring can support the authority and stewardship vision of First Nations across the region; to bolster fisheries and ecosystems through boom and bust cycles of salmon production by reducing harvest in years of low returns and creating opportunities when returns exceed expectations. These efforts will help to maintain long-standing social-ecological linkages between Indigenous peoples and wild salmon, even in the face of climate variability and change.

Artificial Intelligence, Meets Indigenous Weir Fishing Technology

SalmonVision braids computer-vision AI technology with Indigenous fishing weirs (river-spanning fences used to sustainably harvest salmon for millennia) to count returning salmon in real-time. Developing a web application will allow First Nation users to upload their own video and perform computer-vision analysis independently, empowering Indigenous fisheries managers to make real-time decisions regarding harvest in subsistence, commercial, and recreational fisheries. This process is rooted in deep collaboration and co-development, reframing the approach and motivation for development and implementation of AI systems to address needs of Indigenous communities.  

Computer vision, a field of AI focused on automated object recognition and tracking, holds immense promise in ecological monitoring. Until recently access to the power of these emerging AI tools has not been harnessed effectively for salmon conservation and has rarely been co-developed with Indigenous Nations. Since 2020, Wild Salmon Center (WSC) and our research partners have been working to develop computer-vision (CV) models for automated counting and classification of migrating salmon using data from in-river camera weirs, sonar, and aerial drones. We have built deep research partnerships, brought innovation and major new technical capacity to salmon conservation, and amassed more than four million frames of annotated video and sonar data for training computer-vision models.

CV deep learning can enable rapid processing of data, with transformative applications in salmon population assessment and fisheries management.  Working with two First Nations fishery programs in British Columbia, Canada, we trained and tested CV models for object detection and tracking, performing automated video enumeration of salmon passing two First Nation run weirs.  Models were trained with more than 500,000 frames of video data encompassing 12 species, including seven species of anadromous salmonids. Top-performing models achieved a mean average precision (mAP) of 67.6% and species-specific mAP scores > 90% for coho and > 80% for sockeye salmon when trained with a combined dataset of Kitwanga and Bear Rivers’ salmon annotations. 

These efforts have illuminated a deep need in the broader community for the development of an interactive web-based portal enabling users to upload, analyze, visualize, and export salmon count data in a range of formats that can inform timely decision-making. Creating an intuitive, unified space for computer-vision analysis will unlock the potential of CV across salmon conservation, removing previous technical barriers that have limited access to their possibilities. Building in data review, visualization, and annotation tools, will further enhance the value of the web app with continued improvement of CV models by storing reviewed videos as annotated training data.

Current funding supports research, development, and application of CV tools, but with limited budget for preliminary app design; work is underway to design the user experience and app functionality. Project partners Aeria have developed a preliminary web portal with computer vision performing automated detection and classification of salmon from aerial drone surveys, along with spatial data visualization. Leveraging these investments and existing IP, we are seeking resources to work with our partners to develop a beta version of the web application, called SalmonVision.

Current collaborations include work with the Heiltsuk, Kitasoo-Xai’xais, Gitga’at, Gitanyow, Nuxalk, Wuikinuxv, Taku River Tlingit, and Haida Nations, as well as the Skeena Fishery Commission an umbrella fisheries organization representing First Nations in the Skeena River. Each of these Nations and their fishery programs are working to assert their governance authority over salmon fisheries and to provide sustainable fishery benefits for current and future generations. Project lead Dr. William Atlas has worked in close partnership with BC First Nations for more than a decade, and the need for these tools has emerged from conversations and collaborations with each partner Nation. 

The benefits of these tools will flow not only to fishery decision-makers, but to First Nations people in each of the partner communities. By providing access to real-time information on returning salmon numbers, each Nation will manage thier salmon fishery and ensure sustained fishery access and cultural opportunities connected to wild salmon. Many of these communities are extremely remote, and people depend on the annual return of salmon to provide access to affordable and healthy food. Pairing the application of computer-vision tools with the development of small-scale terminal (in-river) fisheries, our work is providing access not only to information, but increasing access to local, healthy, and sustainable food. Further, projects include education and community outreach, typically with a live stream of the project site and in-water cameras for use in the Nations on-reserve schools, and field trips to the project sites where students and community members can learn about their Nation’s efforts to steward salmon to support sustainable ongoing connections to wild salmon. 

On-the-ground projects are run by staff and technicians from each Nation and their allied organizations. These individuals benefit directly from the investments in infrastructure and technology, streamlining their workflow, improving daily operations and data quality, and providing real-time insights that are reflected back to community harvesters on a regular (weekly) basis through a variety of preferred communications outlets (e.g. social media, newsletters, etc). 

Ultimately, this project is about more than the application of AI technology for efficient salmon counting, it is about empowering resurgent Indigenous governance authority with cutting-edge tools and rebuilding place-based relationships to salmon and the healthy, thriving communities that are sustained by those relationships. Our project team has also helped First Nations partners leverage funding and investments in computer-vision technology to secure funding for on-the-ground projects, with over $1.8 million USD in additional direct funding to First Nations communities around BC since 2020.

Our team at SalmonVision Collaborative brings diverse backgrounds, skills, and networks. WSC works around the Pacific Rim supporting Indigenous stewardship of salmon ecosystems. Pacific Salmon Foundation (PSF) is a BC-based eNGO with significant scientific capacity and funding and a large network of First Nations and agency partners. Partners at SFU Computing Science bring expertise in CV modeling, network and computing systems design, and software development, and Aeria is an ecological technology company with alignment in: skills, areas of active research, and values. 

Dr. William Atlas has worked with the Heiltsuk Nation for over 12 years and has worked directly with all of the Central Coast First Nations (CCFN: Wuikinuxv, Nuxalk, Kitasoo Xai’xais, Heiltsuk) since 2019. These relationships are the foundation and north star of our work and are rooted in deep reciprocity, and mutual commitment to outcomes that benefit Indigenous people. In 2020, Dr. Atlas was nominated by the CCFN to serve as their representative on the Pacific Salmon Commission’s Northern Panel, a bilateral panel of Alaskans and British Columbians working to manage salmon fisheries in Southeast Alaska and Northern BC. Relationships with other partner First Nations have grown organically through the PSC table, and regular knowledge-sharing forums convening First Nations fishery managers. 

The need for AI tools that can streamline analysis of video and sonar datasets has been articulated by partner First Nations, and we began work in 2020 on training and testing computer-vision models with partnership and data from the Haida Fishery Program, Gitanyow Fishery Authority, and Skeena Fishery Commission, while simultaneously using and leveraging grant funds to build video weir programs with the Kitasoo Xai’xais and Heiltsuk Nations. Co-development is at the center of our research process; as trust and awareness of our work have grown, First Nations staff have typically reached out to our team about support with local projects in their communities. For example, the Gitga’at Oceans and Lands Department (GOLD) connected with our team in 2022 about co-developing proposals to support a video weir in the Kitkiata River, a traditional village site in their territory. Working with GOLD we wrote proposals to build a program reflecting the Gitga’at stewardship vision in their home watershed and succeeded in securing >$450,000 USD in support. As part of this Indigenous-led project we have provided technical support for project design, attended meetings with hereditary and elected leadership, and worked with GOLD to ensure that community feedback, concerns, and needs are incorporated into project plans. 

This work continues to center the needs and vision of Indigenous partner communities. We recently convened partners and practitioners to provide feedback on preliminary user interface design and functionality of the web app. Having proven the feasibility of computer-vision for salmon identification and counting, we are focused on providing in-season information to First Nations on returning salmon abundance in 2024. We have worked closely with partners to build computer and camera systems for real-time data analysis, and work is already underway to design and engineer a web app for initial testing within 12 months.

Our project is in a growth phase. During the project’s first phase (2020-2022) we tested and proved the potential for computer-vision deep learning to support automated salmon counting and identification, we amassed a substantial library of training data, built local programmatic capacity with First Nations partners, and developed computer, camera and network systems to support real-time monitoring of salmon returns. This work has built trust and enthusiasm in the ability of our team to deliver outcomes for partner First Nations, all while directly or indirectly providing funding for on-the-ground capacity. 

With a second major grant from the BC Salmon Recovery and Innovation Fund secured through our Canadian partner organization the Pacific Salmon Foundation (PSF), we have expanded the project to include 10 First Nation-led monitoring projects, with new and emerging interest from other Indigenous Nations not currently partnered in the work, including in Alaska and the lower 48 states. To meet this demand and grow the impact of this work for current and new partners we need to streamline our workflow and create tools that allow First Nation fishery managers and stewardship practitioners to analyze and interpret their own data using pre-trained computer-vision models. SalmonVision will ultimately be a tool that is available to all salmon managers and conservation practitioners. Building it alongside BC First Nations means that it is tailored to their specific applications, needs, and concerns, and that we are prioritizing successful technology deployment and positive community benefits with the network of partners first and foremost.

For our work to take the next step and to meet its full potential for partner First Nations and Tribes, we need to solve a core technical challenge: how do we put computer-vision deep learning tools in the hands of people without expertise in programming languages such as PYTHON? To meet this challenge we have envisioned SalmonVision, an-all-in-one web-based app and platform where practitioners can complete their entire workflow by uploading their video, sonar, and drone data, run pre-trained computer vision models, interpret and check outputs from these analyses, and generate new annotated training data for iterative model retraining and performance improvements. Since the project started in 2020, we’ve worked with researchers from Simon Fraser University’s (SFU) computing sciences department, under the supervision of Dr. JC Liu. This partnership has been exceedingly fruitful and the graduate students we’ve worked with have brought a high level of excellence and professionalism that has been instrumental to our overall success. However, the current network of partners has maxed our team's ability to directly support these analyses, and we’ve long heard from partners that they want to be able to run and check their own computer-vision analyses to eliminate bottlenecks in the analytical process. 

For SalmonVision to be successful long-term it will require tools that users can navigate easily, and outputs from the computer-vision analysis that users can check and trust. The web app is a critical step for bringing this vision to life. By automating the analysis and data interpretation pipeline we can support a broader constellation of projects, and scale the impact of our work beyond the current 10 partner projects. A new partnership with Canadian-based ecological technology company Aeria will provide some of the backend architecture for the new web app, but much work remains to be done to bring the web app to the stage of a minimum viable product. Ultimately, our collective aim is to provide these tools to First Nations and Tribes free of charge, however, ongoing costs of site maintenance, model training and testing, and backend support for the user community will pose challenges without a product that can also be marketed to other users such as state or federal management agencies, and the private sector. We are seeking funding and expert support to build out the first phase of the SalmonVision web app, and to develop a business plan that allows us to expand our network of collaborators, generate revenue, and provide long-term access to First Nations and Tribal partners free of charge.

Dr. Atlas has well over a decade of experience working directly with and for First Nations communities on the Central Coast of BC, starting with work for the Heiltsuk Nation in 2012. He joined Wild Salmon Center in 2020 and has continued to work closely with First Nations across the North and Central Coast, building a network of relationships and collaborations with Nations around the region by supporting technical capacity, co-development of research, and fundraising. Serving in a trusted role as a technical advisor to the CCFN, and as their representative on the PSC Northern Panel, Dr. Atlas routinely gathers feedback on fishery management priorities from CCFN, and brings these concerns to the Northern Panel for inclusion in Canada’s bilateral fishery management at the PSC. He also regularly reports back to CCFN with updates from the PSC Northern Panel, as well as on funding and collaborative opportunities that may be aligned with each Nation’s vision for salmon stewardship in their territory. In this capacity, our network has grown to include many First Nations in the Northern BC, the Skeena River, and Northern Transboundary watersheds (e.g. Taku River Tlingit). These collaborative relationships are strengthened by regular trips into communities for events such as potlatches, salmon open houses, meetings with tribal council or hereditary leadership, and ongoing co-development of multiple First Nation-led projects around Northern BC. Ultimately, we are guests in the First Nations and Tribal territories where we work, and we understand that trust and partnership are earned through a sustained commitment to partners and the communities they serve.

Investments in the development of AI have grown exponentially in recent years. These tools hold immense potential to transform society, both for the better and with unintended negative consequences, creating an urgent need for human-centered and ethical AI development. In particular, the needs and concerns of Indigenous people have often been overlooked by a technology sector focused on rapid growth, data harvesting, and competition in a profit-driven industry. Among First Nations and Tribes along the North Pacific Rim, no species or natural resource is more consequential than wild Pacific salmon, and Indigenous Nations are increasingly reasserting their authority for governance, and revitalizing stewardship and harvest practices that have been the foundation of sustainable social-ecological systems for millennia. 

Along with bringing innovative AI tools that support Indigenous-led salmon management, our approach to co-development and deployment of AI tools in the SalmonVision collaborative is truly innovative. First Nation partners have been actively engaged and supported with funding to participate in every step of the project, from concept development, in-the-field data collection, and annotation, to publication of research results. These data are owned by each Nation and are used by our project through data-sharing agreements. Respecting the time, knowledge, and sovereignty of Indigenous people in developing these AI tools sets our efforts apart from much of the industry. In addition to building a web app for this project, we are also training edge-capable CV models that will be run in the field to provide timely and accurate counts in real-time, reducing costs and environmental impacts driven by growing cloud storage and computing demands.   

SalmonVision AI will support First Nation and Tribal-led salmon monitoring programs with computer-vision tools for automated analysis of video, sonar, and drone data that are already collected routinely by Indigenous fishery programs. The web application will enable users to upload and analyze their own data with pre-trained models and to conduct data checking and corrections in an efficient and streamlined manner. By working directly with 10 First Nation-led projects, we are building tools catered to the specific needs of partner communities, and providing access to cutting-edge technology and expertise free of charge. There are hundreds of Indigenous and locally-led salmon monitoring programs across the Pacific Northwest and Alaska, however, user-centered AI tools have yet to be developed for salmon stewardship applications. Part of the challenge is the limited profit potential or AI tools developed for niche applications like salmon monitoring, meaning that large technology companies and centers of technological innovation have been slow to take up these opportunities for co-development of AI with Indigenous salmon stewardship practitioners. 

Our solution, SalmonVision AI, is innovative and needed, braiding traditional Indigenous harvest and stewardship technology (e.g. weirs) with thousands of years of history across the Pacific Rim, with cutting-edge computer-vision deep learning models and a web application that puts these AI technologies at the fingertips of Indigenous salmon managers and promotes the revitalization of terminal and selective fisheries, providing broad social-ecological benefits to Indigenous salmon fishing communities.

SalmonVision Collaborative will catalyze transformative change in salmon fishery management and co-governance, empowering Indigenous rights holders with real-time information on the abundance of salmon returning to watersheds within their territory. In the world of fishery management knowledge is power, and government managers often use the excuse that populations are data-limited as justification for the perpetuation of status quo management regimes. Colonial management systems for salmon have done immense harm to Indigenous people and to wild Pacific salmon, robbing governance authority from Indigenous rights holders, outlawing traditional management systems and fishing methods, and contributing to the collapse of wild salmon populations across much of their North Pacific range. 

Reliable and timely monitoring the abundance of returning adult salmon to their natal watersheds is a core need for Indigenous fishery managers, who are increasingly asserting their management authority through co-governance agreements with colonial governments. Population monitoring has traditionally been challenged by the high cost of monitoring programs and limited staff and capacity within rural and remote Indigenous communities, as well as the large financial and time commitment associated with post-season data review. In addition, because of the time required to review data, information is rarely available in-season to guide precautionary fishery management. As climate change accelerates, salmon returns are becoming increasingly unpredictable, heightening the need for timely information that can support fishery management decisions by Indigenous Nations and their federal or state co-governance partners. 

Putting real-time information in their hands, we will empower Indigenous communities as leaders and decision-makers in fishery co-governance, equipping Nations with information that is in many cases superior to what is known by government management agencies. By co-creating a web application and suite of computer-vision tools with Indigenous Nations, we will build a platform that will support real-time data uploading and cloud-based computer-vision analysis from a constellation of First Nations partners monitoring projects, as well as batched uploading post-season for analysis using pre-trained-models. These data will be owned by partner Nations and managed using best practices for Indigenous data sovereignty. 

Braiding these technology tools with community salmon harvest programs and management processes will inform sustainable management of salmon fisheries for the benefit of current and future generations, and will support the ongoing revitalization of Indigenous stewardship and fishing practices. Real-time data will enable managers to reduce harvest impacts in years when low abundance warrants precaution, and to mobilize fishers to reap the benefits of abundance during years of unexpectedly high returns. Thus, the benefits of our proposed work are multifaceted, supporting Indigenous governance authority with technology to deliver reliable real-time data, but also supporting sustained access to salmon fishery opportunities for food security and economic development, by promoting the re-emergence of Indigenous place-based systems of harvest and stewardship for wild salmon.

Goal 1: Empower Indigenous salmon stewards with tools for real-time monitoring of returning wild salmon, supporting Indigenous governance authority in fishery management. 

Indicators:

1. Provide in-season and post-season analysis using computer-vision models to deliver reliable salmon count data for 10 First Nation partner projects during 2024 season.

2. Minimum viable product for SalmonVision web app to support in-season and post-season data review, and new training data annotations by First Nation partners by Q2 2025. 

Goal 2: Strengthen food security in remote Indigenous communities through the integration of AI technology and support for on-the-ground Nation-led salmon stewardship programs, and sustainable salmon harvest opportunities. 

Indicators:

1. Leveraged funding supporting Indigenous-led salmon stewardship programs

2. Harvest opportunities for Indigenous fishers and community members through the SalmonVision collaborative projects

3. Harvest information and management plans co-developed with partner Nations identifying whether harvest and stewardship goals are met.

Goal 3: Life below water: Protect the abundance and biodiversity of wild salmon for future generations, through improvements in the quality and extent of monitoring programs tracking their returns. 

Indicators: 

1. Manage harvest impacts to reduce risks of overfishing and ensure that sufficient numbers of salmon survive to spawn to support ecosystem values and population resilience. 

2. Double network of collaborators and end-users of SalmonVision by 2025 to provide improved monitoring capability and real-time data integration across the North American range of wild Pacific Salmon.

Our proposed project braids multiple technologies, including computer-vision deep learning models, web apps, in-water video cameras, sonar sensors, and aerial drones, with Indigenous harvest management systems that have been used for millennia including weirs; river-spanning fences that were traditionally used for harvest and stewardship of salmon and are a physical embodiment of Indigenous tenure and governance authority. 

Data are collected in the field using three distinct technologies: (1) underwater video cameras mounted in weirs, (2) sonar sensors monitoring a cross-section of river for migrating salmon, and (3) aerial drone surveys to enumerate salmon efficiently during the spawning system. These diverse tools provide Indigenous practitioners with a diverse toolbox of methods for monitoring salmon abundance, depending on their specific watershed context and data needs. CV deep learning models have been developed using open-source models in the YOLO (you only look once) family of models, implemented in PYTHON. These CV models can be run on the edge using microprocessors (Jetson Nano by NVIDIA), or data can be streamed to the cloud for analysis using cloud GPU resources. A motion detection algorithm is applied to the data onsite to reduce the volume of uploads and maximize the efficiency of the data review workflow. 

SalmonVision is a web app that will give users the ability to upload their own data, either automatically via satellite internet connection at their project site, or manually once data is retrieved, and to analyze their data using pre-trained computer-vision models. While CV models and training data libraries have been built, and model performance continues to improve, the app is currently in the design stage. We have therefore focused on the conceptual design of the user interface, identifying what features and functions are needed, and understanding how and when Indigenous partners want access to their results for decision making.

British Columbia, Yukon Territory

Alaska, Washington, Oregon

Our team includes diverse individuals offering unique skill sets to advance the promise of SalmonVision. Three full-time staff at WSC and Canadian partner Pacific Salmon Foundation work on this project, with eight graduate students at SFU computing sciences working on aspects of computer-vision, remote network systems, and app development. A contractor is completing initial app design, and a team of three computer-scientists at Aeria are engaged to develop the app. Most importantly, 32 First Nations fishery managers and technical staff are involved as project collaborators, applying SalmonVision tools in the field and working with our team to iteratively improve tools. 

Initial work on computer-vision models for automated salmon detection and tracking began in 2020, when we received funding from the BC Salmon Innovation and Recovery Fund. This project successfully demonstrated the potential for computer-vision tools in salmon stewardship and fostered expanded collaboration with First Nations across BC. A follow-up grant in 2023 now supports the next phase of this work, which will provide in-season analysis tools to partners beginning in 2024.

Wild Salmon Center has worked with Indigenous and rural communities throughout our history and across our geographies, but institutionalizing that work from within began in earnest in 2018 with a more reflective and intentional assessment of equity issues. Examining both internal WSC structures and external partner organization strategies, as well as the wider social and political context of the United States, we formed a JEDI committee (justice, equity, diversity, and inclusion). Since then, we have had several team and organization-wide conversations about our socio-political positioning as conservationists, a historically white-lensed sector, and the work needed going forward. In 2020, we finalized a JEDI committee charter with objectives and a timeline for both the committee and WSC’s everyday operations. 

The JEDI committee continues to work to synthesize what staff have learned from Indigenous partners and rural communities on British Columbia’s Central Coast, Alaska’s Bristol Bay, Washington’s Olympic Peninsula, and Oregon/California’s Klamath River basin into a framework WSC staff can use to strengthen how we show up for our tribal and rural partners. Vetted by outside equity experts, we internally published this Indigenous Framework to better equip WSC to build equitable, lasting partnerships with local community members and to create a standard of partnership practices where we constantly interrogate our histories and social positioning within the organization. The strategy will also help inform us when we connect with the wider community of NGOs, foundations, and academic partners. The framework states: 

“Our conservation efforts will begin with engagement with local Tribes/First Nations. All WSC staff will instinctively ask, and be able to accurately answer at every critical project/program/campaign juncture, these three pivotal questions:

1. Where are the Tribes/First Nations on this issue?

2. What is the best strategy for furthering thoughtful and respectful engagement?

3. Are we fostering equal opportunities for Tribal leadership, collaboration, and engagement in our joint salmon stronghold work with Tribes?”

Even though we “published” the Framework, this effort is iterative and ongoing, there will be no time when we are “done” building equity, but we continue to center and be led by our Indigenous and rural community partners who have been most impacted by the threats to salmon and their watersheds.

To date, development of computer-vision models, remote network and computing technologies, and deployment of these tools has been supported by grant funding. The ambitious goals set out by the SalmonVision collaborative requires streamlining of data collection, review, and annotation pipelines, to allow us to support a broader and more diverse group of collaborators. We have given numerous presentations on CV salmon counting and SalmonVision tools at conferences and agency meetings, and have already had preliminary discussions with staff from BC Hydro, Pacific Salmon Commission, Fisheries and Oceans Canada, Oregon Department of Fish and Wildlife, Douglas County Public Utility, and the Washington Department of Fish and Wildlife about the potential for SalmonVision to support their analysis and workflow.

Initially, we will provide clients with a web app including three modules for CV analysis of video, sonar, and aerial footage, as well as optional support with camera and computer systems for remote processing and uploading. This app will radically expedite the workflow and efficiency of reviewing these data and producing timely population counts for wild salmon and will support key user functionalities including data review, visualization, and annotation. Technical support for system design and build will further enable users to rapidly adopt and effectively implement CV analysis for salmon counting. These tools are clearly in demand and will be rolled out with some of the aforementioned organizations in a phased approach within 12 months. This revenue stream will support permanent project staff and cover cloud storage and computing costs, and continued development of the tools offered by SalmonVision. It will also allow us to provide the services free of charge for Indigenous and local community organizations. 

We envision continual growth in the tools, services, and modules offered by the SalmonVision collaborative. Around the Pacific Rim, hundreds of salmon monitoring projects generate large quantities of video, sonar, and aerial drone data. Resources are always limited for monitoring, creating a large unmet demand for the tools we are building. Our tools will allow First Nations, local communities, agencies, and private sector partners the ability to analyze more data, rapidly, and at dramatically reduced costs.

Wild Salmon Center and most project partners are not-for-profit, and initial research and development have been supported by grants two major grants from BC SRIF, with additional funding from Experiment.com, Royal Bank of Canada, and Canada’s MITACS graduate research fellowships. Additional co-development of proposals with First Nations partners has leveraged >$1.8 million USD in additional funding to support program infrastructure and capacity. Through this research and local programmatic funding, we have built a network of projects, large quantities of training data, and a replicable system for automated monitoring and counting of wild salmon. 

We envision a significant long-term revenue stream from fee-for-service contracts with agencies, public utilities, and other private sector organizations, and we have begun conversations with a number of agencies and utilities with known needs for automation of their video and sonar analysis workflow. At the moment we remain laser-focused on developing and applying CV and app tools and strengthening collaborative partnerships but recognize that within 12 months we will need to decide on how the SalmonVision collaborative will ultimately be structured to meet financial needs for cost recovery, ongoing investments in tools and services, and revenue sharing. As we move out of the research and testing phase, the core team of collaborators (WSC, PSF, SFU, and Aeria) has agreed to continue these conversations with the aim of having a partnership agreement or spin-off business structure in place before expanding our revenue and client community significantly.