https://phuchoang2603.github.io/sis-arxiv-vad-papers/authors/elisa-riccia/Recent content in Elisa Riccia on sis-arxiv-vad-papersHugo -- gohugo.ioenSun, 01 Jan 2023 00:00:00 +0000https://phuchoang2603.github.io/sis-arxiv-vad-papers/papers/delving-into-clip-latent-space-for-video-anomaly-recognition/Sun, 01 Jan 2023 00:00:00 +0000https://phuchoang2603.github.io/sis-arxiv-vad-papers/papers/delving-into-clip-latent-space-for-video-anomaly-recognition/Proposes AnomalyCLIP, a novel method leveraging Large Language and Vision (LLV) models like CLIP, combined with multiple instance learning and a re-centring transformation of the CLIP feature space, to detect and classify video anomalies and recognize anomaly types. Introduces a Selector model with prompt learning and a Temporal Transformer-based model for temporal dependency modeling; demonstrates state-of-the-art performance on multiple benchmarks.