Convolutional neural network for target recognition and recognition method thereof

The invention discloses a convolutional neural network for target recognition and a recognition method thereof, and relates to the technical field of target recognition. The invention provides a novel hardware implementation mode. Comprising the following steps: for a novel microstructure and a processing process of a convolutional neural network data stream, recording front input data and an output label at an input layer/output layer of the convolutional neural network data stream by adopting a magnetic random access memory; then, comparing new data and existing data on the input layer; and only updating and recalculating subsets that change in event-driven manner. In an application scene with high-intensity requirements or limits in the aspects of complexity/energy/battery life, the event-driven convolutional neural network based on the magnetic random access memory and the recognition method thereof have high-end performance in all directions, such as resolution, working frequency, data throughput and acc