首页>
外国专利>
OFFLINE COMBINATION OF CONVOLUTION/DE-CONVOLUTION LAYER AND BATCH NORMALIZATION LAYER OF CONVOLUTION NEUTRAL NETWORK MODEL USED FOR AUTOMATIC DRIVING VEHICLE
OFFLINE COMBINATION OF CONVOLUTION/DE-CONVOLUTION LAYER AND BATCH NORMALIZATION LAYER OF CONVOLUTION NEUTRAL NETWORK MODEL USED FOR AUTOMATIC DRIVING VEHICLE
展开▼
机译:用于自动驾驶汽车的卷积神经网络模型的卷积/去卷积层和批量归一化层的离线组合
展开▼
页面导航
摘要
著录项
相似文献
摘要
PROBLEM TO BE SOLVED: To provide a system of convolution neutral network (CNN) model of accelerated batch normalization capable of being used in an automatic driving vehicle.SOLUTION: The system extracts a plurality of first layer groups from a first CNN model and each group includes a first convolution layer and a first batch normalization layer. Regarding each group, the system calculates a first scale vector and a first shift vector on the basis of the first batch normalization layer and generates a second convolution layer showing a corresponding group in the plurality of first groups on the basis of the first convolution layer, the first scale vector and the first shift vector. In addition, the system generates an accelerated CNN model on the basis of the second convolution layer corresponding to the plurality of first groups and classifies targets detected by an automatic driving vehicle.SELECTED DRAWING: Figure 9
展开▼