Here's another look into what we've been up to recently.
As reported in the last update, we've been having some trouble getting a reliable video feed out of the new thermal cameras we're using. We've made a significant breakthrough on the software side around packet handling and sync recovery which has improved performance significantly. Despite this, we're still not 100% satisfied with the robustness of the video feed, so we spent Friday afternoon with the project's electronics advisor Roger McKenzie, looking into the electrical aspects of the camera and it's connection to the Raspberry Pi. Although we learned a lot and managed to find some areas of improvement we didn't find a conclusive cause for the problems we've been having. Our work continues...
You can find the little utility we've created to read and validate packets from the Lepton 3 camera on Github.
The API and web interface parts of the Cacophony Project web server (known as "Full Noise") have been split up. This simplifies the code base (there was little shared code between the two parts anyway), making it easier for new contributors to get in to. The change also allows us to more easily deploy the two parts independently (i.e. different DNS names, separate servers). This means that in addition to the pre-existing repository on Github which now just has the API server, we have a new repository for the web interface code.
We're currently testing the newly arrange code and will deploy it to production soon.
Machine Learning Presentation
There was an excellent machine learning presentation at last week's Christchurch Python meetup by Matthew Aitchison which covered using Python and TensorFlow. The talk was an excellent opportunity for some of our team to brush up on machine learning, and also to generate interest in the Cacophony Project with some of the attendees. There may be some exciting news to report soon as a result ...
Our team has started to use the Rocket.Chat messaging system for real-time communication and are keen to offer this as another way for interested people to get in touch with the team. In order to join us, point your web browser at https://chat.nzoss.nz/ and register an account. Once logged in, join us at the #cacophony channel. As well as a web interface, Rocket.Chat has open source desktop and mobile apps if you prefer those.