I have been making good progress on the balloons tracking, the speed is good, but it is not as reliable as I would like it to be. A friend of ours who studies computer vision gave me a few pointers, and an alternative solution : Caffe, Berkeley’s framework of computer vision deep learning. It allows the installation of trained neural networks, via the Model Zoo. I am looking into it, but since I do not know anything in this field I am a bit lost.
We started drafting specs for the camera with Perceval, we probably will work more on this in the coming week.