Invertibility of Convolutional Generative Networks from Partial Measurements

By Fangchang Ma, Ulas Ayaz, Sertac Karaman. 2018

70 citations recorded by the source corpus.

Subfields: Generative Models, Statistical Learning Theory

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Generative Models

In Statistical Learning Theory

Citations in the corpus

Interpreting the Latent Space of GANs for Semantic Face Editing (2019, methodology): “To make GANs applicable for real image processing, existing methods proposed to reverse the mapping from the latent space to the image space [27, 39, 24, 5, 16] or learn an additional encoder associated with the GAN training [13, 12, 38].”
Image Quality Assessment: Unifying Structure and Texture Similarity (2020, background): “[50] proved that under Gaussian-distributed random weights and ReLU nonlinearity, a two-layer CNN is injective provided that it is sufficiently expansive (i.”
Deep Neural Networks are Surprisingly Reversible: A Baseline for Zero-Shot Inversion (2021, background): “Besides discriminative models, [21, 53] investigate the theoretical invertibility of the deep generative model.”
InOut: Diverse Image Outpainting via GAN Inversion (2021, methodology): “Optimization-based meth-ods [1,2,7,19,21] use gradient-based optimization methods with reconstruction loss as the objective function to ﬁnd latent codes that can recover input images.”
Global Guarantees for Enforcing Deep Generative Priors by Empirical Risk (2017, result): “The results of this paper, while directly stated for fully connected nets have already been extended to the case of convolutional generative nets, as proven in [72].”
Unsupervised 3D Shape Completion through GAN Inversion (2021, background): “Typically, the latent vector can be optimized based on gradient descent [22, 20], or projected by an extra encoder from the image space [39, 17].”
Relative gradient optimization of the Jacobian term in unsupervised deep learning (2020, background): “[38] Fangchang Ma, Ulas Ayaz, and Sertac Karaman.”
InterFaceGAN: Interpreting the Disentangled Face Representation Learned by GANs (2020, methodology): “Prior work either performed instance-level optimization [46], [47], [48] or explicitly learned an encoder to reverse the generator [49], [50], [51].”
Image Processing Using Multi-Code GAN Prior (2019, methodology): “Some work theoretically explored the prior provided by deep generative models [32, 18], but the results using GAN prior to real image processing are still unsatisfying.”
From Continuity to Editability: Inverting GANs with Consecutive Images (2021, background): “The former individually optimizes the latent code for a specific image, with a concentration on pixel-wise reconstruction [1, 2, 3, 13].”
In-Domain GAN Inversion for Real Image Editing (2020, methodology): “in a deterministic model [23,31]. The other is optimization-based, which deals with a single instance at one time by directly optimizing the latent code to minimize the pixel-wise reconstruction loss [19,7,21,24]. Some work combines these two ideas by using the encoder to generate an initialization ”
HyperInverter: Improving StyleGAN Inversion via Hypernetwork (2021, methodology): “There are two general approaches to determining a latent code of a GAN model given an input image: by an iterative optimization [1, 10, 21, 25, 32] and by inference with an encoder [35, 37, 44, 58].”