Sharks, crocodiles and mud make understanding the effect of commercial fishing on local marine life a difficult task in northern Australia.
But a new artificial intelligence trial in Darwin is helping speed up the process and free up scientists to do more important work.
Northern Territory Fisheries have started teaching an artificial neural network how to identify various fish species in Darwin Harbour.
The harbour is home to about 415 known species of marine life, including rays and dugongs, but data collection a challenge for scientists who have to encounter a mix of crocodiles, sharks, murky water, weather and tides of up to eight metres.
Analysing underwater video frame-by-frame to categorise fish can take humans up to three hours, while the software can take as little as 10-to-20 minutes.
"At the moment we're focused on two-to-15 species of fish but the model can only do as much as you provide in terms of a training data set," NT Fisheries acting principal scientist Shane Penny told AAP.
"Our focus is on those commercially-important species but more broadly, our research has to take into account the whole ecosystem."
Dr Penny said the new software would allow his small team of three scientists to work remotely and process raw video quickly so more time could be spent on analysis or more data collection.
That could include increasing the number of species monitored and whether fishing some species was impacting diversity or populations in the harbour.
Microsoft says the project has already piqued the interest of fisheries departments across Australia and could even be used to monitor land species such as the kookaburra.