Anomaly and activity recognition in a video surveillance using Masked Autoencoder

Video cameras are used to monitor many of the public locations in today's globe, including supermarkets, public gardens, malls, university campuses, and so on. There is a requirement to provide crucial security and monitor aberrant anomaly activity at such locations. The fundamental disadvantage of the previous method is that manual procedures must be conducted 24 hours a day, seven days a week, and human error is a possibility. The goal of this study is to detect anomalies and recognise human activity in films. The masked spatiotemporal autoencoder (MAE) is used to detect video anomalies. The suggested model design has three basic elements: a spatial encoder, a temporal encoder-decoder, and a spatial decoder.