A CONVOLUTIONAL NEURAL NETWORK (CNN) SYSTEM BASED ON RESOLUTION-LIMITED SMALL-SCALE CNN MODULES

摘要

Embodiments of a convolutional neutral network (CNN) system based on usingresolution-limited small-scale CNN modules are disclosed. In some embodiments,a CNNsystem includes: a receiving module for receiving an input image of a firstimage size, thereceiving module can be used to partition the input image into a set ofsubimages of a secondimage size; a first processing stage that includes a first hardware CNN moduleconfiguredwith a maximum input image size, the first hardware CNN module is configuredtosequentially receive the set of subimages and sequentially process thereceived subimages togenerate a set of outputs; a merging module for merging the sets of outputsinto a set ofmerged feature maps; and a second processing stage for receiving the set offeature maps andprocessing the set of feature maps to generate an output including at leastone prediction onthe input image.