Ah, I had missed that. I guess this will mitigate the risk a lot, although I would still like to have seen results against a test set of images from a different

Ah, I had missed that. I guess this will mitigate the risk a lot, although I would still like to have seen results against a test set of images from a different context (social media for example).