More AI improvements

A group from Auckland University has been training AI models using our thermal video library.

We were contacted early last year by a group at Auckland University looking for some data to apply some of the latest AI tools to.  We happily supplied them with access to our thermal video library (with associated tags) and left them to it.  We recently received a report on their work and we'd like to share their promising results with you. The team have used some of the latest tools and have achieved much better results than the previous models we have been using. Attached is the report with all the details.

Species Classification in Thermal Imaging Videos.pdf

Note: this is just a report summarising the team's research - it has not yet been peer reviewed or published elsewhere.

We are very pleased with the results the team have achieved and we will be including the details of these models in our upcoming software updates for the Cacophony monitoring tools.  We're confident this will help increase the accuracy of our classifications and thus reduce effort even further.

We see the real value of the open source approach when talented folks contribute like this. Many thanks to Hamish Duncanson, Charles Tremlett, Saahil Deshpande, Yahui Cai and Goutham Menon!

As an open source project, everything The Cacophony Project does is openly available.  If you think our data can help progress your research, feel free to get in touch.

As always, we welcome your feedback so don't hesitate to get in touch - leave a comment below or email us at blog@cacophony.org.nz.