I’ll look it up later but if you look at any of the recent computer vision conferences and search for things like anomaly detection with either or attention or lstm networks you should be able to find stuff.
But those papers (the ones I'm familiar with) are human labelled - the challenge with anomaly detection is winnowing out the events that humans believe are really anomalies from the coincidences and permutations in the data which normal statistics sees as very unusual.
Maybe I'm mistaken but from what I remember I had read a couple that were not human labeled. Let me take a gander later today.
Thank you - I'm really interested as anomaly detection is such a pain!