首页>
外国专利>
GANGenerative Adversarial Network LEARNING METHOD AND LEARNING DEVICE FOR REDUCING DISTORTION OCCURRED IN WARPED IMAGE GENERATED IN PROCESS OF STABILIZING JITTERED IMAGE BY USING GAN TO ENHANCE FAULT TOLERANCE AND FLUCTUATION ROBUSTNESS IN EXTREME SITUATIONS
GANGenerative Adversarial Network LEARNING METHOD AND LEARNING DEVICE FOR REDUCING DISTORTION OCCURRED IN WARPED IMAGE GENERATED IN PROCESS OF STABILIZING JITTERED IMAGE BY USING GAN TO ENHANCE FAULT TOLERANCE AND FLUCTUATION ROBUSTNESS IN EXTREME SITUATIONS
The present invention reduces distortion generated in a warped image using a Generative Adversarial Network (GAN), which is provided to improve Fault Tolerance and Fluctuation Robustness in extreme situations. It relates to a learning method for (a) when an initial image is obtained, an adjustment layer included in a generating network causes at least some initial feature value (Initial Feature Value) converting the initial image into an Adjusted Image by adjusting the ; and (b) obtaining at least a portion of (i) a Naturality Score, (ii) a Maintenance Score, and (iii) a Similarity Score, the loss layer included in the generating network (Loss Layer), with reference to the naturalness score, the characteristic maintenance score, and the similarity score to generate a generating network loss (Generating Network Loss) to learn the parameters of the generating network; In addition, the present invention may be used for behavior prediction, ultra-precise object detection or tracking, and the like.
展开▼