Research Article

Anomaly Detection in Moving Crowds through Spatiotemporal Autoencoding and Additional Attention

Figure 5

Illustration of pixel-level differences in the Ped2 dataset: (a) input data, (b) pixel-level differences of ConvLSTM-AE, (c) pixel-level differences of Conv-AE, and (d) pixel-level differences of our approach.
(a)
(b)
(c)
(d)