Hybrid CNN–SRU/LSTM with multiple instance learning for real-time video anomaly detection in surveillance

Key Points

• Anomaly detection is accomplished in real-time using a hybrid cnn-sru/lstm model, enhancing surveillance efficiency.
• The detection system successfully identifies anomalies within video feeds, providing immediate feedback to security operators.
• Assessment utilized multiple instance learning to train the model across varying surveillance video datasets to improve accuracy.
• This model may enable more responsive security systems, although external validation is necessary for widespread implementation.