CMU’s O-VAD detects industrial video anomalies by tracking object state over time
CarnegieMellonU · hf · 2026-07-28
CMU researchers present O-VAD, a training-free framework for industrial video anomaly detection.
The method is designed for settings where generic VLM-based anomaly detection performs poorly because industrial scenes have strict physical and procedural constraints. Instead of retraining on normal clips or injecting domain knowledge, O-VAD:
- tracks object-level spatial and temporal state changes over time
- reasons over object-wise trajectories to identify abnormal objects in grounded frames
- outputs interpretable reports about the anomaly process and anomaly type
The authors report results on three IVAD datasets, where O-VAD outperforms frontier VLMs, agentic frameworks, and traditional VAD methods fine-tuned on the same datasets.
More from Research
- NeurIPS 2026 MATH-AI workshop opens papers on agentic mathematical reasoning — KaiyuYang4 · 2026-07-28
- Different coding agents can produce very different estimates from the same model spec — JessicaHullman · 2026-07-28
- Progressive context extension may suit linear-attention hybrids better than short-SWA models — stochasticchasm · 2026-07-28
- Google DeepMind says LLMs can prove theorems but still cannot make scientific “jumps” — kylekabasares · 2026-07-28
- Latent Action Model talk will show how to learn Super Mario from observation alone — ceciletamura · 2026-07-28
- GLM 5 follows up with per-head Muon to balance attention-head updates — stochasticchasm · 2026-07-28