Anomaly Detection in Videos Using Two-Stream Autoencoder with Post Hoc Interpretability

Figure 3

Temporal model architecture. The final output is the reconstructed frame sequence.