ML Seminar: Abnormal event detection in training/exercise videos – University of Copenhagen

ML Seminar: Abnormal event detection in training/exercise videos

ML Seminar: Talk by Huamin Ren

Abstract
An algorithm to learn a behavior-specific dictionary based on sparse coding is used for representing normal patterns in training videos. Anomalities are detected based on their large reconstruction error. Experiments are shown on one public dataset and one dataset which we recorded and are about to publish.