Aligned Disentangling Generative AdversarialNetwork (AD-GAN) introduces representationdisentanglement to separate content representation from style representation . With this framework, spatial structure can be preserved explicitly, enabling asignificant reduction of macro-level lossy transformation . AD-GAN leads to significant improvement over the current best unsupervised methods by an average 17.8% relatively (w.r.t. themetric DICE) on four cell nuclei datasets . We also propose anovel training algorithm able to align the disentangled content in the latentspace to reduce micro-level lossesy transformation. Evaluations on real-world 2Dand 3D datasets show that AD-GA substantially outperforms the other comparisonmethods and the professional software both quantitatively and qualitatively. As an unsupervisory method, AD-GRAPHIC DICE has performed competitive with the best supervised models, according to the authors . The authors also propose a training algorithm to .
Author(s) : Kai Yao, Kaizhu Huang, Jie Sun, Curran JudeLinks : PDF - Abstract
Code :
Keywords : ad - training - gan - nuclei - level -