首页>
外国专利>
A method and a learning device for learning an object detector capable of CNN-based hardware optimization using image concatenation for long-distance detection or military purposes, a test method and a testing device using the same, {LEARNING METHOD AND LEARNING DEVICE FOR OBJECT DETECTOR WITH HARDWARE OPTIMIZATION ON BASED ON CNN FOR DE DETECTION AT DISCANCED OR MILEDGING INDEPENDING CONTACT
A method and a learning device for learning an object detector capable of CNN-based hardware optimization using image concatenation for long-distance detection or military purposes, a test method and a testing device using the same, {LEARNING METHOD AND LEARNING DEVICE FOR OBJECT DETECTOR WITH HARDWARE OPTIMIZATION ON BASED ON CNN FOR DE DETECTION AT DISCANCED OR MILEDGING INDEPENDING CONTACT
PROBLEM TO BE SOLVED: To provide a method for learning an object detector capable of hardware optimization based on CNN based on image concatenation for long-range detection or military purpose. The CNN may be redesigned if the size of the object changes as the resolution and the focal length of the KPI changes. The method comprises: (a) concatenating an nth processed image corresponding to an nth target region; (b) a first to nth object in the nth processed image using an integrated feature map with RPN. Generating a proposal and applying a pooling operation to a region corresponding to the first to nth object proposals on the integrated pitcher map with the pooling layer; (c) outputting with the FC loss layer to the FC layer Referring to the object detection information, acquiring the first to nth FC losses; [Selection diagram] Figure 2
展开▼
