Trained Still Wins: Narrowing the Gap to Zero-Shot Video Anomaly Detection
A simplified future-frame-prediction network is trained on the normal-only training split of UCSD Ped1, UCSD Ped2, and CUHK Avenue, reaching a mean AUC of 0.848, and a prompt-sensitivity analysis shows how much of the training-free result depends on prompt wording, cases where a more sophisticated method did not beat a...