A synthesizable Verilog design for two staged simplified convolutional neural network arithmetic. System generates 8 output vectors which can be used to classify objects.
- Two parametrized architectures are developed with one being throughput oriented and other being minimal area oriented.