In which Wildbook did the issue occur? Whiskerbook
What operating system were you using? MacOS 11.6.1
What web browser were you using? Chrome
What is your role on the site? Researcher
What happened? It looks like during the automated annotations, the view point is being mispredicted. This came up in a prior post of mine, linked below. I am seeing a handful of images in my data with this issue. Below are two examples. As recommended in the prior post, I am going to go through and manually redraw the annotations for the images, but I wanted to call attention to the problem, since I am seeing it quite a few times.
I went through the Jhalana Forest Reserve, and it looks like the large majority of viewpoints are correct. This is the cross-application of a model originally trained on African data, and that may minorly impact this, but generally this looks to be within the normal range of % viewpoint errors. It did seem like backright viewpoints were more often confused for left than other errors I saw. We’ll watch for this more and see if a future model retrain is needed.