LEARNING METHOD AND LEARNING DEVICE FOR SWITCHING MODES OF AUTONOMOUS VEHICLE BASED ON ON-DEVICE STANDALONE PREDICTION TO THEREBY ACHIEVE SAFETY OF AUTONOMOUS DRIVING AND TESTING METHOD AND TESTING DEVICE USING THE SAME

摘要

A learning method for generating a parameter that can indicate reliability of object detection during a process of object detection, comprising: (a) causing, by a learning apparatus, a convolutional layer to generate a convolutional feature map by applying a convolution operation to a training image; (b) causing the learning device to generate an RPN confidence map including an RPN confidence score by the anchor layer; (c) causing the learning apparatus to generate a CNN confidence map by causing the FC layer to generate a CNN confidence score; and (d) the learning apparatus causes the loss layer to use the RPN loss and CNN loss generated by referring to the RPN confidence map, the CNN confidence map, the prediction object detection result and the GT object detection result. Backpropagation (backpropagation) Including; learning at least some of the parameters of the CNN and the RPN by performing.