TL;DR使用自监督特征和对抗训练,提出了一种无监督目标分割方法 MOVE。该方法在多个评估数据集上实现了无监督显著对象检测和分割的 SotA 表现,单目标发现的 CorLoc 平均提高了 7.2%,无监督类别不可知的对象检测的相对 AP 平均提高了 53%。
Abstract
We introduce MOVE, a novel method to segment objects without any form of
supervision. MOVE exploits the fact that foreground objects can be shifted
locally relative to their initial position and result in realistic
(undistorted) new images. This property allows us to train a segmentation model
on a dataset of images without annotation and to achieve state of