Machine vision model for detecting when fishing activity is occurring in on-board camera footage

The Fisheries NZ on-board cameras programme captures video footage from ~220 inshore commercial fishing vessels. This footage is a key source of information that supports data-driven and responsive fisheries management. MPI only requires high-resolution footage when fishing activity is occurring – transmitting high-resolution footage outside of these times increases transmission, storage, and processing costs. The camera system uses automated human detection to act as a proxy for fishing activity. However, human presence in footage does not perfectly correlate with fishing activity. You will mainly work with a team from MPI and Spark to investigate and trial new machine vision approaches that more accurately identify when fishing activity is occurring. The primary working location will be the MPI offices in Wellington. The information that you will work with may be commercially sensitive. You will need to abide by MPI's Code of Conduct and information security and privacy policy.