2040 Newsletter

Kia ora,

Welcome to spring and te wiki o te reo Māori.  I'm currently learning te reo Māori at Te Wānanga o Aotearoa (level 2) and am loving it. I encourage any of you considering this to give it a go.

We have some exciting updates to share with you in this newsletter, across all of our product range, including:

• Thermal cameras: research about when they are most cost effective, their accuracy and their ability to detect Dama wallabies. As well as various software improvements, including adding photos to stations, hierarchical tags and recognising penguins
• Bird monitors: software updates that allow you to synchronise bird monitors and change the length of the recordings, as well as being able to upload from SD cards.
• Traps  - funding from 2050, progress on auto-reset, intelligent triggering, and auto kill, along with updates on trials by DOC and Predator Free Banks Peninsula.

We have been in the news with articles from Predator Free NZ about the history and goals of The Cacophony Project and about using AI to reach the Predator Free 2050 goal, along with a video journalism project about Predator Free 2050.

In other industry news, it's been great to see the progress NZ Auto traps are making. I recommend reading the AT220 Field Trial report from the Bay of Islands. It's great to see images of dead possums like the one below which emphasise the importance of scent trails.

Finally I'd like to thank the organisers of the SONZi conference which I had the pleasure of attending in Nelson last month. It was fantastic to hear about the great work that is happening around the NZ sanctuaries and to meet everyone.

Kā mihi nui,
Shaun Ryan
2040 Limited.

Thermal camera updates

Research, improvements and the sale continues.

The following research published about our thermal cameras over the past few months:

• The Department of Conservation in conjunction with The Cacophony project have published a paper comparing thermal cameras and trail cameras. This looks at the total cost of ownership and recommends different monitoring scenarios that are suited to each type of camera.
• Cacophony published some research on their blog looking at the accuracy of the thermal camera machine vision. The summary is that, the closer an animal is the higher the accuracy and smaller animals need to be closer. For example, the accuracy is close to 100% when a possum is 6-8m away, and for rats when they're 3-4m away.
• MPI commissioned 2040 to investigate the accuracy of the thermal cameras for detecting Dama wallabies with help from the Bay of Plenty Regional Council. The summary is that the accuracy is reasonably good (~57%) when the wallabies are close to the camera, and poor (~34%) when they are a long way from the camera. These should improve with more training data.

We've continued to make software improvements to the cameras:

• You can now add photos to your stations (camera locations). This is really useful for trying to interpret a thermal video if you're not familiar with the location. 
• We have added hierarchical tagging which allows you to tag with more granularity - for example, you can tag rats or mice rather than rodents. This will probably be released later today.
• The camera can now automatically identify penguins. This was created from a relatively small training set, mostly from captive penguins. It may not be that accurate on penguins in the wild, but this will improve as we get more data.
• You can now change your display name and must log in with your email address. This is helpful when you're trying to identify who is doing tagging. Some customers have multiple people tagging and this can help identify some that may need extra training.
• We have added a section to the thermal camera manual describing the email alerts and how these can be set up. This lets you get a near real time alert when a predator is automatically detected by the camera.
• We are still having a camera sale

Station picture example

Bird monitor updates

Software updates. Bird taggers wanted.

We have made some updates to our bird monitor software.

• You can now synchronise multiple bird monitors so they all record at the same time.
• We have also released an update that allows you to change the length of the recordings.
• You can also manually upload bird recordings from the SD card (2nd part of this post). This means for remote monitors you can just swap SD cards, rather than bring the monitor to reception, upload and return.

We are still looking for people willing to tag bird recordings. This will help us automatically identify these birds.  Please contact us if you are interested in doing this.

Trap update

The Cacophony Project has received funding from Predator Free 2050 to fund further development of our trap. This includes an automated reset, intelligent trigger and an auto-kill mechanism.

The Cacophony Project have been testing the 2nd version of their automated reset mechanism in the field. They call this the spool trigger. This is lighter, smaller and cheaper to manufacture than our previous, trigger mechanism. They have done a reasonable about of testing for durability and more recently have been confirming that the mechanism is fast enough to catch predators. Last week the trap caught a couple of cats and a possum. Yay!

The Cacophony Project recently published a summary of their research into intelligent triggers for the trap. They are currently testing IR cameras mounted inside the trap. A recent experiment using a thermal camera (outside of the trap) to activate the trap when a target was identified failed because the predators were coming from the side of the trap where they were too far away to identify reliably and the computation may have been too slow. Having cameras inside the trap will help address the first of these problems and we are investigating more efficient algorithms to address the latter. For those interested, a random forest algorithm is showing some great promise, thanks to open source contributor Jamie Heather from Carbon Critical.

The Cacophony Project have also been developing an auto kill mechanism. This has been really interesting. After capturing the animal we are able to create an environment where we can position them precisely. We've found that our self resetting DOC 250 won't kill a possum instantly. We have had more success with adding a spike to this and hitting the possum on the back of the neck (see the image below). The aim is to provide a mechanism that will kill all predators instantly and will be NAWAC approved.

Finally, a quick update on 2 trials of our traps, both in the same region.  DOC have been using 10 traps at Te Waihora. A recent article describes how their Cacophony traps have a 10x higher catch rates than traditional traps used in the same area. Predator Free Banks Peninsula have also been trialling 10 traps on Kaitorete, which has been inconclusive at best. The main reason for this has been a lack of target predators as detected by thermal cameras on each of the traps. They are planning to test the traps in a different location where target predators have been detected.

We'll keep you up to date with the results of our development, as this continues.

2040 on social media

2040 is on Facebook, Twitter, Instagram and LinkedIn Follow us wherever you are to stay up to date.

Sign up to this newsletter

Have you been forwarded this newsletter?

Sign up here to receive this newsletter directly. We aim to send out one per month but often end up doing it less frequently because life gets in the way.

2040 Ltd is a social enterprise whose mission is to eradicate predators from NZ by manufacturing and selling open source technology developed by The Cacophony Project.