首页>
外国专利>
METHOD FOR ACQUIRING SAMPLE IMAGES FOR INSPECTING LABEL AMONG AUTO-LABELED IMAGES TO BE USED FOR LEARNING OF NEURAL NETWORK AND SAMPLE IMAGE ACQUIRING DEVICE USING THE SAME
METHOD FOR ACQUIRING SAMPLE IMAGES FOR INSPECTING LABEL AMONG AUTO-LABELED IMAGES TO BE USED FOR LEARNING OF NEURAL NETWORK AND SAMPLE IMAGE ACQUIRING DEVICE USING THE SAME
The present invention is a method for acquiring a sample image for label inspection among auto-labeled images for deep learning network learning while optimizing a manual labeling sampling process and reducing an annotation cost, wherein the sample image acquisition device comprises: And the second image, the convolution layer to generate the first feature map and the second feature map, and the pooling layer to generate the first pooled feature map and the second pooled feature map, and concatenate Causing the generated feature map to be generated; Allowing a deep learning classifier to obtain the converted feature map and generate class information; And calculating a probability of an abnormal class element of an abnormal class group, determining whether the auto-labeled image is a differential image, and selecting the auto-labeled image as the sample image for label inspection. Features are provided. In addition, the method may be performed using a robust algorithm having a plurality of modified pairs. In addition, the present invention can more accurately detect a dangerous situation.
展开▼
