2022-06-07 更新
Differentiable Gaussianization Layers for Inverse Problems Regularized by Deep Generative Models
Authors:Dongzhuo Li
Deep generative models such as GANs and normalizing flows are powerful priors. They can regularize inverse problems to reduce ill-posedness and attain high-quality results. However, the latent vector of such deep generative models can fall out of the desired high-dimensional standard Gaussian distribution during an inversion, particularly in the presence of noise in data or inaccurate forward models. In such a case, deep generative models are ineffective in attaining high-fidelity solutions. To address this issue, we propose to reparameterize and Gaussianize the latent vector using novel differentiable data-dependent layers wherein custom operators are defined by solving optimization problems. These proposed layers constrain an inversion to find feasible in-distribution solutions. We tested and validated our technique on three inversion tasks: compressive-sensing MRI, image deblurring, and eikonal tomography (a nonlinear PDE-constrained inverse problem), using two representative deep generative models: StyleGAN2 and Glow, and achieved state-of-the-art results.
PDF 26 pages, 15 figures, 9 tables
论文截图
Learning Robust Representations Of Generative Models Using Set-Based Artificial Fingerprints
Authors:Hae Jin Song, Wael AbdAlmageed
With recent progress in deep generative models, the problem of identifying synthetic data and comparing their underlying generative processes has become an imperative task for various reasons, including fighting visual misinformation and source attribution. Existing methods often approximate the distance between the models via their sample distributions. In this paper, we approach the problem of fingerprinting generative models by learning representations that encode the residual artifacts left by the generative models as unique signals that identify the source models. We consider these unique traces (a.k.a. “artificial fingerprints”) as representations of generative models, and demonstrate their usefulness in both the discriminative task of source attribution and the unsupervised task of defining a similarity between the underlying models. We first extend the existing studies on fingerprints of GANs to four representative classes of generative models (VAEs, Flows, GANs and score-based models), and demonstrate their existence and attributability. We then improve the stability and attributability of the fingerprints by proposing a new learning method based on set-encoding and contrastive training. Our set-encoder, unlike existing methods that operate on individual images, learns fingerprints from a \textit{set} of images. We demonstrate improvements in the stability and attributability through comparisons to state-of-the-art fingerprint methods and ablation studies. Further, our method employs contrastive training to learn an implicit similarity between models. We discover latent families of generative models using this metric in a standard hierarchical clustering algorithm.
PDF
论文截图
Blended Latent Diffusion
Authors:Omri Avrahami, Ohad Fried, Dani Lischinski
The tremendous progress in neural image generation, coupled with the emergence of seemingly omnipotent vision-language models has finally enabled text-based interfaces for creating and editing images. Handling generic images requires a diverse underlying generative model, hence the latest works utilize diffusion models, which were shown to surpass GANs in terms of diversity. One major drawback of diffusion models, however, is their relatively slow inference time. In this paper, we present an accelerated solution to the task of local text-driven editing of generic images, where the desired edits are confined to a user-provided mask. Our solution leverages a recent text-to-image Latent Diffusion Model (LDM), which speeds up diffusion by operating in a lower-dimensional latent space. We first convert the LDM into a local image editor by incorporating Blended Diffusion into it. Next we propose an optimization-based solution for the inherent inability of this LDM to accurately reconstruct images. Finally, we address the scenario of performing local edits using thin masks. We evaluate our method against the available baselines both qualitatively and quantitatively and demonstrate that in addition to being faster, our method achieves better precision than the baselines while mitigating some of their artifacts. Project page is available at https://omriavrahami.com/blended-latent-diffusion-page/
PDF Project page is available at https://omriavrahami.com/blended-latent-diffusion-page/
论文截图
PixelFolder: An Efficient Progressive Pixel Synthesis Network for Image Generation
Authors:Jing He, Yiyi Zhou, Qi Zhang, Jun Peng, Yunhang Shen, Xiaoshuai Sun, Chao Chen, Rongrong Ji
Pixel synthesis is a promising research paradigm for image generation, which can well exploit pixel-wise prior knowledge for generation. However, existing methods still suffer from excessive memory footprint and computation overhead. In this paper, we propose a progressive pixel synthesis network towards efficient image generation, coined as PixelFolder. Specifically, PixelFolder formulates image generation as a progressive pixel regression problem and synthesizes images by a multi-stage paradigm, which can greatly reduce the overhead caused by large tensor transformations. In addition, we introduce novel pixel folding operations to further improve model efficiency while maintaining pixel-wise prior knowledge for end-to-end regression. With these innovative designs, we greatly reduce the expenditure of pixel synthesis, e.g., reducing 90% computation and 57% parameters compared to the latest pixel synthesis method called CIPS. To validate our approach, we conduct extensive experiments on two benchmark datasets, namely FFHQ and LSUN Church. The experimental results show that with much less expenditure, PixelFolder obtains new state-of-the-art (SOTA) performance on two benchmark datasets, i.e., 3.77 FID and 2.45 FID on FFHQ and LSUN Church, respectively. Meanwhile, PixelFolder is also more efficient than the SOTA methods like StyleGAN2, reducing about 74% computation and 36% parameters, respectively. These results greatly validate the effectiveness of the proposed PixelFolder.
PDF 11 pages, 7 figures
论文截图
ContraCLIP: Interpretable GAN generation driven by pairs of contrasting sentences
Authors:Christos Tzelepis, James Oldfield, Georgios Tzimiropoulos, Ioannis Patras
This work addresses the problem of discovering non-linear interpretable paths in the latent space of pre-trained GANs in a model-agnostic manner. In the proposed method, the discovery is driven by a set of pairs of natural language sentences with contrasting semantics, named semantic dipoles, that serve as the limits of the interpretation that we require by the trainable latent paths to encode. By using the pre-trained CLIP encoder, the sentences are projected into the vision-language space, where they serve as dipoles, and where RBF-based warping functions define a set of non-linear directional paths, one for each semantic dipole, allowing in this way traversals from one semantic pole to the other. By defining an objective that discovers paths in the latent space of GANs that generate changes along the desired paths in the vision-language embedding space, we provide an intuitive way of controlling the underlying generative factors and address some of the limitations of the state-of-the-art works, namely, that a) they are typically tailored to specific GAN architectures (i.e., StyleGAN), b) they disregard the relative position of the manipulated and the original image in the image embedding and the relative position of the image and the text embeddings, and c) they lead to abrupt image manipulations and quickly arrive at regions of low density and, thus, low image quality, providing limited control of the generative factors. We provide extensive qualitative and quantitative results that demonstrate our claims with two pre-trained GANs, and make the code and the pre-trained models publicly available at: https://github.com/chi0tzp/ContraCLIP
PDF
论文截图
Rethinking Positive Sampling for Contrastive Learning with Kernel
Authors:Benoit Dufumier, Carlo Alberto Barbano, Robin Louiset, Edouard Duchesnay, Pietro Gori
Data augmentation is a crucial component in unsupervised contrastive learning (CL). It determines how positive samples are defined and, ultimately, the quality of the representation. While efficient augmentations have been found for standard vision datasets, such as ImageNet, it is still an open problem in other applications, such as medical imaging, or in datasets with easy-to-learn but irrelevant imaging features. In this work, we propose a new way to define positive samples using kernel theory along with a novel loss called decoupled uniformity. We propose to integrate prior information, learnt from generative models or given as auxiliary attributes, into contrastive learning, to make it less dependent on data augmentation. We draw a connection between contrastive learning and the conditional mean embedding theory to derive tight bounds on the downstream classification loss. In an unsupervised setting, we empirically demonstrate that CL benefits from generative models, such as VAE and GAN, to less rely on data augmentations. We validate our framework on vision datasets including CIFAR10, CIFAR100, STL10 and ImageNet100 and a brain MRI dataset. In the weakly supervised setting, we demonstrate that our formulation provides state-of-the-art results.
PDF
论文截图