Recent GAN-based architectures have been able to deliver impressive performance on the general task of image-to-image translation. In particular, it was shown that a wide variety of image translation operators may be learned from two image sets, containing images from two different domains, without establishing an explicit pairing between the images. This was made possible by introducing clever regularizers to overcome the under-constrained nature of the unpaired translation problem. In this work, we introduce a novel architecture for unpaired image translation, and explore several new regularizes enabled by it. Specifically, our architecture comprises a pair of GANs, as well as a pair of translators between their respective latent spaces. These cross-translators enable us to impose several regularizing constraints on the learnt image translation operator, collectively referred to as latent cross-consistency. Our results show that our proposed architecture and latent cross-consistency constraints are able to outperform the existing state-of-the-art on a variety of image translation tasks.

The emergence of deep generative models has recently enabled the automatic generation of massive amounts of graphical content, both in 2D and in 3D. Generative Adversarial Networks (GANs) and style control mechanisms, such as Adaptive Instance Normalization (AdaIN), have proved particularly effective in this context, culminating in the state-of-the-art StyleGAN architecture. While such models are able to learn diverse distributions, provided a sufficiently large training set, they are not well-suited for scenarios where the distribution of the training data exhibits a multi-modal behavior. In such cases, reshaping a uniform or normal distribution over the latent space into a complex multi-modal distribution in the data domain is challenging, and the generator might fail to sample the target distribution well. Furthermore, existing unsupervised generative models are not able to control the mode of the generated samples independently of the other visual attributes, despite the fact that they are typically disentangled in the training data. In this paper, we introduce uMM-GAN, a novel architecture designed to better model multi-modal distributions, in an unsupervised fashion. Building upon the StyleGAN architecture, our network learns multiple modes, in a completely unsupervised manner , and combines them using a set of learned weights. We demonstrate that this approach is capable of effectively approximating a complex distribution as a superposition of multiple simple ones. We further show that uMM-GAN effectively disentangles between modes and style, thereby providing an independent degree of control over the generated content.

Video frame interpolation is a long-studied problem in the video processing field. Recently, deep learning approaches have been applied to this problem, showing impressive results on low-resolution benchmarks. However, these methods do not scale-up favorably to high resolutions. Specifically, when the motion exceeds a typical number of pixels, their interpolation quality is degraded. Moreover, their run time renders them impractical for real-time applications. In this paper we propose IM-Net: an interpolated motion neural network. We use an economic structured architecture and end-to-end training with multi-scale tailored losses. In particular, we formulate interpolated motion estimation as classification rather than regression. IM-Net outperforms previous methods by more than 1.3dB (PSNR) on a high resolution version of the recently introduced Vimeo triplet dataset. Moreover, the network runs in less than 33msec on a single GPU for HD resolution.

Abstract We present a method for determining which facial parts (mouth, nose, etc.) best characterize an individual, given a set of that individual's portraits. We introduce a novel distinctiveness analysis of a set of portraits, which leverages the deep features extracted by a pre‐trained face recognition CNN and a hair segmentation FCN, in the context of a weakly supervised metric learning scheme. Our analysis enables the generation of a polarized class activation map (PCAM) for an individual's portrait via a transformation that localizes and amplifies the discriminative regions of the deep feature maps extracted by the aforementioned networks. A user study that we conducted shows that there is a surprisingly good agreement between the face parts that users indicate as characteristic and the face parts automatically selected by our method. We demonstrate a few applications of our method, including determining the most and the least representative portraits among a set of portraits of an individual, and the creation of facial hybrids: portraits that combine the characteristic recognizable facial features of two individuals. Our face characterization analysis is also effective for ranking portraits in order to find an individual's look‐alikes (Doppelgängers).

Example-based texture synthesis has been an active research problem for over two decades. Still, synthesizing textures with nonlocal structures remains a challenge. In this article, we present a texture synthesis technique that builds upon convolutional neural networks and extracted statistics of pretrained deep features. We introduce a structural energy, based on correlations among deep features, which capture the self-similarities and regularities characterizing the texture. Specifically, we show that our technique can synthesize textures that have structures of various scales, local and nonlocal, and the combination of the two.