Anomaly-Led Prompting Learning Caption Generating Model and Benchmark
·12528 words·59 mins
Qianyue Bao
,
Fang Liu
,
Licheng Jiao
,
Yang Liu
,
Shuo Li
,
Lingling Li
,
Xu Liu
,
Xinyi Wang
,
Baoliang Chen
Introduces a new task for comprehensive video anomaly captioning, proposes a large-scale benchmark dataset CVACBench with fine-grained annotations, and designs a baseline model AGPFormer using prompt learning to improve anomaly understanding and description accuracy.