Found 1 document(s)
Page 1 of 1
-
Vision-language anomaly detectors describe normal and abnormal appearance with text, and their gains are commonly credited to the semantic content of these descriptions. This attribution is plausible in the zero-shot set...