EvAn: Neuromorphic Event-Based Sparse Anomaly Detection

Event-based cameras are bio-inspired novel sensors that asynchronously record changes in illumination the form of events. This principle results significant advantages over conventional cameras, such as low power utilization, high dynamic range, and no motion blur. Moreover, by design, encode only relative between scene sensor not static background to yield a very sparse data structure. In this paper, we leverage these an event camera toward critical vision application—video anomaly detection. We propose detection solution domain with conditional Generative Adversarial Network (cGAN) made up submanifold convolution layers. Video analytics tasks depend on history at each pixel. To enable this, also put forward generic unsupervised deep learning learn memory surface known Deep Learning (DL) surface. DL encodes temporal information readily available from while retaining sparsity data. Since there is existing dataset for domain, provide set anomalies. empirically validate our architecture, composed convolutional layers, proposed online dataset. Careful analysis network reveals presented method massive reduction computational complexity good performance compared previous state-of-the-art frame-based networks.