This project initially started with the goal of creating a tool to dramatically improve monitoring how the environment responds to predator eradication. In New Zealand we are lucky that the bird song is literally the canary in the mine:
Firstly, we love the government’s new announcement of the target to be predator free by 2050. Bold, ballsy and just the sort of thing New Zealand should do. What we would like to do here is a little analysis of why we think it will happen sooner than that.
Moore's law is the observation that, over the history of computing hardware, the number of transistors in a dense integrated circuit doubles approximately every two years. This has morphed into a remarkably consistent observation that Information Technology either doubles in performance or halves in cost about every two years
Trapping, poison, hunting, and fences have all been turned against pests to help prevent the ongoing slaughter of New Zealand birds. However, we know that current trapping methods aren’t cutting it. As part of our experimenting we have had video trail cameras pointing at existing (traditional) traps to measure how well different digital lures work. What has completely surprised us is that as little as 1% of the time that a possum turned up on camera a trap was activated.
The goal of this project is to develop tools that eliminate 100% of predators. To do this the device must therefore be able to detect 100% of predators. Chew cards and tracking tunnels can miss over 60% of predators. Standard camera traps are thought to miss as little as 5% of predators due to not starting fast enough or the light/sound scaring animals away. There are also issues with false positives making it difficult and time consuming to filter the videos
By Brent Martin, adjunct senior research fellow, University of Canterbury
The aim of this project was to see if the latest Machine Learning (Artificial Intelligence) tools could correctly identify the difference between rats, stoats, possums and others from videos collected from the field. This project was given to a set of 28 final year honours students at University of Canterbury.
Summary of the technical features of the project and how we are progressing with each.
We have started testing digital lures with a camera to observe the different rates of animal interactions. These tests are aimed at possums for the simple reason that our property is infested with them. Our initial impressions are...
Machine learning (a branch of Artificial Intelligence) is the general term for how computers learn to recognise patterns. Machine Learning has become very powerful with the advent of inexpensive computer processing and the explosion in available data enabled by the internet and cloud computing. Capitalising on all this data, deep neural networks have recently shown order-of-magnitude improvements in tasks such as automatic speech recognition and computer vision.
This project humbly began by doing a bit of local pest eradication on our small property in Akaroa. After a couple of years I’m sure the bird volume has gone up, thanks to the Akaroa community all doing their bit. I thought this was cool but how could I know for sure? From investigating the manual methods it seemed like computers would be a great way to measure bird volume and record trends over time.
The Cacophony Project has been founded on the idea that some problems are too big, too complex (and too important) for any one commercial entity to solve. For a really huge problem like restoring NZ's native ecosystems, innovation is needed from all over the place. We recognise that a few experts can contribute what they know at low cost to themselves but at high value to the rest of us.
The Cacophony Project was featured on the TV3 news station Newshub...
We start our first blog with what sounds like an outrageously bold claim; that it is possible to make eradication of predators in New Zealand 80,000 times more efficient. Technology could make it possible...