Overview 제목: GANSpace: Discovering Interpretable GAN Controls 저자: Erik Härkönen, Aaron Hertzmann, Jaakko Lehtinen, Sylvain Paris 기관: Aalto Univ., Adobe Research, NVIDIA 학회: ECCV 2020 under review 요약: "Unsupervised identification of interpretable directions in an existing GAN" Pre-trained StyleGAN, BigGAN의 초반 activation space에서 PCA를 수행하여 얻은 각 component를 조절하여 image manipulation을 수행 PCA로 GAN의 laten..

Skip-GANomaly: Skip Connected and Adversarially Trained Encoder-Decoder Anomaly Detection, 2019. Authors: Samet Akçay, Amir Atapour-Abarghouei, Toby P. Breckon Conference: 2019 International Joint Conference on Neural Networks (IJCNN) Github: Link Summary: an U-net like encoder-decoder convolutional neural network with skip-connections and utilized adversarial training scheme the role of skip co..

GANomaly Semi-supervised anomaly detection via adversarial training (2018) Authors: Samet Akcay, Amir Atapour-Abarghouei , and Toby P. Breckon Conference: Asian Conference on Computer Vision (622--637) Organization: Springer Summary: Semi-supervised anomaly detection Architecture: Conditional GAN an adverasarial autoencoder within an Encoder-Decoder-Encoder pipeline Anomalies are detected when t..

f-AnoGAN Authors : Thomas Schleg,Philipp Seeböck, Sebastian M.Waldstein, Georg Langs,Ursula Schmidt-Erfurth Journal : Medical Image Analysis (Impact Factor: 11.148 in 2020) Summary : 기존 AnoGAN가 inference할 때 iterative optimization 의 느린 속도를 개선하고자 encoder 기반의 구조를 제안하였다. 또한, 다양한 encoder 학습 방법을 제안하였다. Architecture : Wasserstein GAN + Encoder Encoder based Anomaly Detection Encoder enables A fast lear..

Analyzing and Improving the Image Quality of StyleGAN Introduction 제목: Analyzing and Improving the Image Quality of StyleGAN 저자: Tero Karras, Samuli Laine, Miika Aittala, Janne Hellsten, Jaakko Lehtinen, Timo Aila 기관: NVIDIA 요약: StyleGAN의 문제를 분석하고, 모델의 구조와 학습 방법을 개선 Generator의 구조 개선 Redesign generator normalization: "Droplet artifacts" 문제 해결 Revisit progressive growing: "phase" artifacts, Shift-..

StyleGAN: A Style-Based Generator Architecture for GANs We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g.,..