Pondering the other day, would like to build a squirrel feeder that will let red squirrels in but not Grey squirrels. Problem, although one is mainly red and the other mainly grey each has a wide variation in colour which can cause confusion even with experienced observers. The most consistent way to distinguish is the tail, a grey has white tips to it's tail hairs giving a white halo whereas the red's tail hair is constant colour along it's length.
Would it be possible to program a Pi to do this? I'm wondering if existing face recognition algorithms would work if given squirrel photos to train on or would it need a bespoke algorithm?