Rethinking Open-World Video Anomaly Detection: Diagnosing Definition Blindness
cs.CV
Submitted: 2026-07-22
Updated: 2026-09-26
Terminology
Sources
- Qwen3-VL Technical Report
- Qwen2.5-VL Technical Report
- No Need For Real Anomaly: MLLM Empowered Zero-Shot Video Anomaly Detection
- Language-guided Open-world Video Anomaly Detection under Weak Supervision
- ESOM: Efficiently Understanding Streaming Video Anomalies with Open-world Dynamic Definitions
- Bounding-Box Trajectories Matter for Video Anomaly Detection
- Video Question Answering for People with Visual Impairments Using an Egocentric 360-Degree Camera
- Instance-Aligned Captions for Explainable Video Anomaly Detection
- InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models