R-CNN LEARNING METHOD AND TESTING METHOD OF OBJECT DETECTOR TO BE USED FOR SURVEILLANCE BASED ON R-CNN CAPABLE OF CONVERTING MODES ACCORDING TO ASPECT RATIOS OR SCALES OF OBJECTS AND LEARNING DEVICE AND TESTING DEVICE USING THE SAME

摘要

The present invention relates to a method for learning an object detector based on R-CNN (Region-based Convolutional Neural Network). Can be determined according to the same characteristics, the learning method, the learning device, the RPN to generate an ROI candidate; causing the pooling layer to output a feature vector; Learning the FC layer and the convolution layer through backpropagation; characterized in that it includes, in this method, the pooling process is distance information and object information obtained through radar, lidar (Lidar) or other sensors A method characterized in that it can be performed according to the actual proportion and actual size of the object by using it, and the learning method and the test method have a similar size from the same viewpoint at a specific location, so that it can be used for monitoring is provided.