Salmon, an important part of Canada’s marine system and food supply, migrate upstream every year in the western province of British Columbia (BC). And every year First Nations groups manually count and monitor the number of fish that pass through—sometimes covering areas larger than the state of Connecticut with just a handful of people.
This count is a crucial part of maintaining healthy salmon stocks in BC, but getting accurate numbers is notoriously tricky due to a number of environmental factors.
That’s where AI comes in:
AI programs are being successfully trained to recognize and count different salmon species based on videos of salmon swimming up the rivers.
- Many species can already be recognized with 70%-90% accuracy.
To make the program more accurate, traditional Indigenous weirs were built to create narrow passages for fish to pass through, making video capture much more efficient.
- Weirs are “fence-like structures built across rivers with a small passage for the fish to pass through.” Source
Now, instead of watching salmon swim by for months, workers can focus on moving other important projects forward.
By 2025, the tool will be implemented by at least 10 First Nations to help monitor and manage salmon populations in BC.