Table 10 Multimodal variational autoencoder (MVAE) for data fusion.

From: Design of an integrated model with temporal graph attention and transformer-augmented RNNs for enhanced anomaly detection

Time (s)

Latent representation (Video)

Latent representation (Audio)

Latent representation (Motion)

Reconstruction error

Anomaly indicator

0–10

0.12

0.15

0.18

0.05

Normal

10–20

0.13

0.17

0.16

0.07

Normal

20–30

0.40

0.38

0.45

0.25

Anomalous

30–40

0.11

0.13

0.14

0.04

Normal

40–50

0.14

0.16

0.15

0.06

Normal

Back to article page