MULTI-STAGE NETWORK FOR TASK-ORIENTED DEEP NEURONAL NETWORK

摘要

This disclosure provides a multi-level network for a task-oriented deep neural network. A method for controlling an autonomous vehicle is disclosed. The method may include receiving image data. The image data can be logistically classified as belonging to a situation from a multiplicity of situations by a first neural network. Based on this logistic classification, the image data can be assigned to a second neural network that is specially trained to deal with a situation. The second neural network can perform a regression of the image data. The vehicle can then be controlled using a command based on the regression.