SYSTEMS AND METHODS FOR ROAD SIGN DETECTION FROM STREET LEVEL IMAGERY USING A MULTI-STAGE NEURAL NETWORK

A road sign detection method comprises obtaining an unlabeled image of a road segment and executing a neural network trained to identify road signs within the unlabeled image. The neural network comprises a first model configured to localize an object in the unlabeled image. The neural network may further comprise a second model configured to generate classification data associated with the cropped image, based on content of the cropped image, where the classification data corresponds to one or more appearance-based attributes of the cropped image and also to determine a category of the cropped image, based on the classification data of the cropped image. The neural network may comprise a third model configured to generate road sign label data based on the determined category of the cropped image, where the road sign label data represents the identified road sign. The method further comprises outputting the road sign label data.