2019年12月13日に「Southern Taiwan University of Science and Technology（南台科技大学）」で開催された「18th International Symposium on Advanced technology (ISAT-18)」において、大学院電気・電子工学専攻修士2年の馬力さん（画像情報メディア研究室・陣キュウ教授）が、研究題目「Adversarial Knowledge Distillation Algorithm for a Compact Generator」の研究発表で「Excellent Oral Paper Presentation Award」を受賞しました。

Many NN structure werer used in haze removal for better result in recently. However most of them cost lots of GPU source in training process. In our proposed method, we handled this proble, woth replacing Batch Norm with Group Norm in shortcut connection part of Unet. Meanwhile we add a logcosh loss, 2 2nd order differentiable function, in our total loss. The experiments shows this method is efficient reduce GPU cost and get better result at same time.