We propose a neuromimetic architecture that can perform always-on pattern recognition. To achieve this, we have extended an existing event-based algorithm (Lagorce et al., 2017), which introduced novel spatio-temporal features as a Hierarchy Of Time-Surfaces (HOTS). Built from asynchronous events captured by a neuromorphic camera, these time surfaces allow to encode the local dynamics of a visual scene and to create an efficient event-based pattern recognition architecture. Inspired by neuroscience, we have extended this method to improve its performance. First, we add a homeostatic gain control on the activity of neurons to improve the learning of spatio-temporal patterns (Grimaldi et al., 2021). We also provide a new mathematical formalism that allows an analogy to be drawn between the HOTS algorithm and Spiking Neural Networks (SNN). Following this analogy, we transform the offline pattern categorization method into an online and event-driven layer. This classifier uses the spiking output of the network to define new time surfaces and we then perform the online classification with a neuromimetic implementation of a multinomial logistic regression. These improvements not only consistently increase the performance of the network, but also bring this event-driven pattern recognition algorithm fully online. The results have been validated on different datasets: PokerDVS (Serrano-Gotarredona and Linares-Barranco, 2015), N-MNIST (Orchard et al., 2015a) and DVS Gesture (Amir et al., 2017). This demonstrates the efficiency of this bio-realistic SNN for ultra-fast object categorization through an event-by-event decision making process.

Convolutional Neural Networks have been considered the go-to option for object recognition in computer vision for the last couple of years. However, their invariance to object’s translations is still deemed as a weak point and remains limited to small translations only via their max-pooling layers. One bio-inspired approach considers the What/Where pathway separation in Mammals to overcome this limitation. This approach works as a nature-inspired attention mechanism, another classical approach of which is Spatial Transformers. These allow an adaptive endto-end learning of different classes of spatial transformations throughout training. In this work, we overview Spatial Transformers as an attention-only mechanism and compare them with the What/Where model. We show that the use of attention restricted or “Foveated” Spatial Transformer Networks, coupled alongside a curriculum learning training scheme and an efficient log-polar visual space entry, provides better performance when compared to the What/Where model, all this without the need for any extra supervision whatsoever.