Skip to main content
. 2024 Mar 28;18:1335422. doi: 10.3389/fnins.2024.1335422

Figure 4.

Figure 4

An illustration of the event-driven depth-first convolution on SENECA. We show a fused layer combining 3×3 convolution and 2×2 max-pooling. The layer processes input events sequentially from the sorted input event queue. Based on the location of the input, events are generated from spatial locations that have been fully updated, and their corresponding memory spaces are released. The channel dimension is omitted for simplicity.