Latent Code Research Articles

Unified cross-modality feature disentangler for unsupervised multi-domain MRI abdomen organs segmentation.

Our contribution is a unified cross-modality feature disentagling approach for multi-domain image translation and multiple organ segmentation. Using CT as the labeled source domain, our approach learns to segment multi-modal (T1-weighted and T2-weighted) MRI having no labeled data. Our approach uses a variational auto-encoder (VAE) to disentangle the image content from style. The VAE constrains the style feature encoding to match a universal prior (Gaussian) that is assumed to span the styles of all the source and target modalities. The extracted image style is converted into a latent style scaling code, which modulates the generator to produce multi-modality images according to the target domain code from the image content features. Finally, we introduce a joint distribution matching discriminator that combines the translated images with task-relevant segmentation probability maps to further constrain and regularize image-to-image (I2I) translations. We performed extensive comparisons to multiple state-of-the-art I2I translation and segmentation methods. Our approach resulted in the lowest average multi-domain image reconstruction error of 1.34±0.04. Our approach produced an average Dice similarity coefficient (DSC) of 0.85 for T1w and 0.90 for T2w MRI for multi-organ segmentation, which was highly comparable to a fully supervised MRI multi-organ segmentation network (DSC of 0.86 for T1w and 0.90 for T2w MRI).

BSD-GAN: Branched Generative Adversarial Network for Scale-Disentangled Representation Learning and Image Synthesis.

We introduce BSD-GAN, a novel multi-branch and scale-disentangled training method which enables unconditional Generative Adversarial Networks (GANs) to learn image representations at multiple scales, benefiting a wide range of generation and editing tasks. The key feature of BSD-GAN is that it is trained in multiple branches, progressively covering both the breadth and depth of the network, as resolutions of the training images increase to reveal finer-scale features. Specifically, each noise vector, as input to the generator network of BSD-GAN, is deliberately split into several sub-vectors, each corresponding to, and is trained to learn, image representations at a particular scale. During training, we progressively "de-freeze" the sub-vectors, one at a time, as a new set of higher-resolution images is employed for training and more network layers are added. A consequence of such an explicit sub-vector designation is that we can directly manipulate and even combine latent (sub-vector) codes which model different feature scales. Extensive experiments demonstrate the effectiveness of our training method in scale-disentangled learning of image representations and synthesis of novel image contents, without any extra labels and without compromising quality of the synthesized high-resolution images. We further demonstrate several image generation and manipulation applications enabled or improved by BSD-GAN.

Generating Realistic Smoke Images With Controllable Smoke Components

Smoke image generation is an important method to solve sample categories imbalance in smoke detection with applications to forest smoke detection, smoky vehicles detection, etc. This article presents two Controllable Smoke Image Generation Neural Networks (CSGNet and CSGNet-v2). More specifically, CSGNet is proposed to generate various smoke images by integrating a smoke component control module (SCM) and a smoke image synthesis module (ISM) with a multi-scale strategy. By changing specified smoke components latent codes, we can generate smoke images with specified smoke components. By fine-tuning smoke components latent codes in SCM, we can fine-tune the smoke components in generated smoke images. To further improve CSGNet, CSGNet-v2 is proposed to make generated smoke images more realistic by adding a smoke image fine-tuning module (IFM). The experiment results show that our methods achieve by far the best results in generating smoke images with controllable smoke components.

GANCCRobot: Generative adversarial nets based chinese calligraphy robot

Robotic calligraphy, as a typical application of robot movement planning, is of great significance for the inheritance and education of calligraphy culture. The existing implementations of such robots often suffer from its limited ability for font generation and evaluation, leading to poor writing style diversity and writing quality. This paper proposes a calligraphic robotic framework based on the generative adversarial nets (GAN) to address such limitation. The robot implemented using such framework is able to learn to write fundamental Chinese character strokes with rich diversities and good quality that is close to the human level, without the requirement of specifically designed evaluation functions thanks to the employment of the revised GAN. In particular, the type information of the stroke is introduced as condition information, and the latent codes are applied to maximize the style quality of the generated strokes. Experimental results demonstrate that the proposed model enables a calligraphic robot to successfully write fundamental Chinese strokes based on a given type and style, with overall good quality. Although the proposed model was evaluated in this report using calligraphy writing, the underpinning research is readily applicable to many other applications, such as robotic graffiti and character style conversion.

Deep Unsupervised Learning of 3D Point Clouds via Graph Topology Inference and Filtering.

We propose a deep autoencoder with graph topology inference and filtering to achieve compact representations of unorganized 3D point clouds in an unsupervised manner. Many previous works discretize 3D points to voxels and then use lattice-based methods to process and learn 3D spatial information; however, this leads to inevitable discretization errors. In this work, we try to handle raw 3D points without such compromise. The proposed networks follow the autoencoder framework with a focus on designing the decoder. The encoder of the proposed networks adopts similar architectures as in PointNet, which is a well-acknowledged method for supervised learning of 3D point clouds. The decoder of the proposed networks involves three novel modules: the folding module, the graph-topology-inference module, and the graph-filtering module. The folding module folds a canonical 2D lattice to the underlying surface of a 3D point cloud, achieving coarse reconstruction; the graph-topology-inference module learns a graph topology to represent pairwise relationships between 3D points, pushing the latent code to preserve both coordinates and pairwise relationships of points in 3D point clouds; and the graph-filtering module couples the above two modules, refining the coarse reconstruction through a learnt graph topology to obtain the final reconstruction. The proposed decoder leverages a learnable graph topology to push the codeword to preserve representative features and further improve the unsupervised-learning performance. We further provide theoretical analyses of the proposed architecture. We provide an upper bound for the reconstruction loss and further show the superiority of graph smoothness over spatial smoothness as a prior to model 3D point clouds. In the experiments, we validate the proposed networks in three tasks, including 3D point cloud reconstruction, visualization, and transfer classification. The experimental results show that (1) the proposed networks outperform the state-of-the-art methods in various tasks, including reconstruction and transfer classification; (2) a graph topology can be inferred as auxiliary information without specific supervision on graph topology inference; (3) graph filtering refines the reconstruction, leading to better performances; and (4) designing a powerful decoder could improve the unsupervised-learning performance, just like a powerful encoder.

A Vector Quantized Variational Autoencoder (VQ-VAE) Autoregressive Neural $F_0$ Model for Statistical Parametric Speech Synthesis

Recurrent neural networks (RNNs) can predict fundamental frequency (F0) for statistical parametric speech synthesis systems, given linguistic features as input. However, these models assume conditional independence between consecutive $F_0$ values, given the RNN state. In a previous study, we proposed autoregressive (AR) neural $F_0$ models to capture the causal dependency of successive $F_0$ values. In subjective evaluations, a deep AR model (DAR) outperformed an RNN. Here, we propose a Vector Quantized Variational Autoencoder (VQ-VAE) neural $F_0$ model that is both more efficient and more interpretable than the DAR. This model has two stages: one uses the VQ-VAE framework to learn a latent code for the $F_0$ contour of each linguistic unit, and other learns to map from linguistic features to latent codes. In contrast to the DAR and RNN, which process the input linguistic features frame-by-frame, the new model converts one linguistic feature vector into one latent code for each linguistic unit. The new model achieves better objective scores than the DAR, has a smaller memory footprint and is computationally faster. Visualization of the latent codes for phones and moras reveals that each latent code represents an $F_0$ shape for a linguistic unit.

Animating landscape

Automatic generation of a high-quality video from a single image remains a challenging task despite the recent advances in deep generative models. This paper proposes a method that can create a high-resolution, long-term animation using convolutional neural networks (CNNs) from a single landscape image where we mainly focus on skies and waters. Our key observation is that the motion (e.g., moving clouds) and appearance (e.g., time-varying colors in the sky) in natural scenes have different time scales. We thus learn them separately and predict them with decoupled control while handling future uncertainty in both predictions by introducing latent codes. Unlike previous methods that infer output frames directly, our CNNs predict spatially-smooth intermediate data, i.e., for motion, flow fields for warping, and for appearance, color transfer maps, via self-supervised learning, i.e., without explicitly-provided ground truth. These intermediate data are applied not to each previous output frame, but to the input image only once for each output frame. This design is crucial to alleviate error accumulation in long-term predictions, which is the essential problem in previous recurrent approaches. The output frames can be looped like cinemagraph, and also be controlled directly by specifying latent codes or indirectly via visual annotations. We demonstrate the effectiveness of our method through comparisons with the state-of-the-arts on video prediction as well as appearance manipulation. Resultant videos, codes, and datasets will be available at http://www.cgg.cs.tsukuba.ac.jp/~endo/projects/AnimatingLandscape.

Generative Models for Low-Dimensional Video Representation and Reconstruction

Finding compact representation of videos is an essential component in almost every problem related to video processing or understanding. In this paper, we propose a generative model to learn compact latent codes that can efficiently represent and reconstruct a video sequence from its missing or under-sampled measurements. We use a generative network that is trained to map a compact code into an image. We first demonstrate that if a video sequence belongs to the range of the pretrained generative network, then we can recover it by estimating the underlying compact latent codes. Then we demonstrate that even if the video sequence does not belong to the range of a pretrained network, we can still recover the true video sequence by jointly updating the latent codes and the weights of the generative network. To avoid overfitting in our model, we regularize the recovery problem by imposing low-rank and similarity constraints on the latent codes of the neighboring frames in the video sequence. We use our methods to recover a variety of videos from compressive measurements at different compression rates. We also demonstrate that we can generate missing frames in a video sequence by interpolating the latent codes of the observed frames in the low-dimensional space.

DUAL-GLOW: Conditional Flow-Based Generative Model for Modality Transfer.

Positron emission tomography (PET) imaging is an imaging modality for diagnosing a number of neurological diseases. In contrast to Magnetic Resonance Imaging (MRI), PET is costly and involves injecting a radioactive substance into the patient. Motivated by developments in modality transfer in vision, we study the generation of certain types of PET images from MRI data. We derive new flow-based generative models which we show perform well in this small sample size regime (much smaller than dataset sizes available in standard vision tasks). Our formulation, DUAL-GLOW, is based on two invertible networks and a relation network that maps the latent spaces to each other. We discuss how given the prior distribution, learning the conditional distribution of PET given the MRI image reduces to obtaining the conditional distribution between the two latent codes w.r.t. the two image types. We also extend our framework to leverage "side" information (or attributes) when available. By controlling the PET generation through "conditioning" on age, our model is also able to capture brain FDG-PET (hypometabolism) changes, as a function of age. We present experiments on the Alzheimers Disease Neuroimaging Initiative (ADNI) dataset with 826 subjects, and obtain good performance in PET image synthesis, qualitatively and quantitatively better than recent works.

Learning Graph Embedding With Adversarial Training Methods.

Graph embedding aims to transfer a graph into vectors to facilitate subsequent graph-analytics tasks like link prediction and graph clustering. Most approaches on graph embedding focus on preserving the graph structure or minimizing the reconstruction errors for graph data. They have mostly overlooked the embedding distribution of the latent codes, which unfortunately may lead to inferior representation in many cases. In this article, we present a novel adversarially regularized framework for graph embedding. By employing the graph convolutional network as an encoder, our framework embeds the topological information and node content into a vector representation, from which a graph decoder is further built to reconstruct the input graph. The adversarial training principle is applied to enforce our latent codes to match a prior Gaussian or uniform distribution. Based on this framework, we derive two variants of the adversarial models, the adversarially regularized graph autoencoder (ARGA) and its variational version, and adversarially regularized variational graph autoencoder (ARVGA), to learn the graph embedding effectively. We also exploit other potential variations of ARGA and ARVGA to get a deeper understanding of our designs. Experimental results that compared 12 algorithms for link prediction and 20 algorithms for graph clustering validate our solutions.

Adversarially Approximated Autoencoder for Image Generation and Manipulation

Regularized autoencoders learn the latent codes, a structure with the regularization under the distribution, which enables them the capability to infer the latent codes given observations and generate new samples given the codes. However, they are sometimes ambiguous as they tend to produce reconstructions that are not necessarily a faithful reproduction of the inputs. The main reason is to enforce the learned latent code distribution to match a prior distribution while the true distribution remains unknown. To improve the reconstruction quality and learn the latent space a manifold structure, this paper presents a novel approach using the adversarially approximated autoencoder (AAAE) to investigate the latent codes with adversarial approximation. Instead of regularizing the latent codes by penalizing on the distance between the distributions of the model and the target, AAAE learns the autoencoder flexibly and approximates the latent space with a simpler generator. The ratio is estimated using a generative adversarial network to enforce the similarity of the distributions. In addition, the image space is regularized with an additional adversarial regularizer. The proposed approach unifies two deep generative models for both latent space inference and diverse generation. The learning scheme is realized without regularization on the latent codes, which also encourages faithful reconstruction. Extensive validation experiments on four real-world datasets demonstrate the superior performance of AAAE. In comparison to the state-of-the-art approaches, AAAE generates samples with better quality and shares the properties of a regularized autoencoder with a nice latent manifold structure.

Hyperprior Induced Unsupervised Disentanglement of Latent Representations

We address the problem of unsupervised disentanglement of latent representations learnt via deep generative models. In contrast to current approaches that operate on the evidence lower bound (ELBO), we argue that statistical independence in the latent space of VAEs can be enforced in a principled hierarchical Bayesian manner. To this effect, we augment the standard VAE with an inverse-Wishart (IW) prior on the covariance matrix of the latent code. By tuning the IW parameters, we are able to encourage (or discourage) independence in the learnt latent dimensions. Extensive experimental results on a range of datasets (2DShapes, 3DChairs, 3DFaces and CelebA) show our approach to outperform the β-VAE and is competitive with the state-of-the-art FactorVAE. Our approach achieves significantly better disentanglement and reconstruction on a new dataset (CorrelatedEllipses) which introduces correlations between the factors of variation.

Understanding VAEs in Fisher-Shannon Plane

In information theory, Fisher information and Shannon information (entropy) are respectively used to quantify the uncertainty associated with the distribution modeling and the uncertainty in specifying the outcome of given variables. These two quantities are complementary and are jointly applied to information behavior analysis in most cases. The uncertainty property in information asserts a fundamental trade-off between Fisher information and Shannon information, which enlightens us the relationship between the encoder and the decoder in variational auto-encoders (VAEs). In this paper, we investigate VAEs in the Fisher-Shannon plane, and demonstrate that the representation learning and the log-likelihood estimation are intrinsically related to these two information quantities. Through extensive qualitative and quantitative experiments, we provide with a better comprehension of VAEs in tasks such as high-resolution reconstruction, and representation learning in the perspective of Fisher information and Shannon information. We further propose a variant of VAEs, termed as Fisher auto-encoder (FAE), for practical needs to balance Fisher information and Shannon information. Our experimental results have demonstrated its promise in improving the reconstruction accuracy and avoiding the noninformative latent code as occurred in previous works.

Learning Disentangled Representation with Pairwise Independence

Unsupervised disentangled representation learning is one of the foundational methods to learn interpretable factors in the data. Existing learning methods are based on the assumption that disentangled factors are mutually independent and incorporate this assumption with the evidence lower bound. However, our experiment reveals that factors in real-world data tend to be pairwise independent. Accordingly, we propose a new method based on a pairwise independence assumption to learn the disentangled representation. The evidence lower bound implicitly encourages mutual independence of latent codes so it is too strong for our assumption. Therefore, we introduce another lower bound in our method. Extensive experiments show that our proposed method gives competitive performances as compared with other state-of-the-art methods.

SAGNet

We present SAGNet, a structure-aware generative model for 3D shapes. Given a set of segmented objects of a certain class, the geometry of their parts and the pairwise relationships between them (the structure) are jointly learned and embedded in a latent space by an autoencoder. The encoder intertwines the geometry and structure features into a single latent code, while the decoder disentangles the features and reconstructs the geometry and structure of the 3D model. Our autoencoder consists of two branches, one for the structure and one for the geometry. The key idea is that during the analysis, the two branches exchange information between them, thereby learning the dependencies between structure and geometry and encoding two augmented features, which are then fused into a single latent code. This explicit intertwining of information enables separately controlling the geometry and the structure of the generated models. We evaluate the performance of our method and conduct an ablation study. We explicitly show that encoding of shapes accounts for both similarities in structure and geometry. A variety of quality results generated by SAGNet are presented.

Reconstructing faces from fMRI patterns using deep generative neural\xa0networks

Although distinct categories are reliably decoded from fMRI brain responses, it has proved more difficult to distinguish visually similar inputs, such as different faces. Here, we apply a recently developed deep learning system to reconstruct face images from human fMRI. We trained a variational auto-encoder (VAE) neural network using a GAN (Generative Adversarial Network) unsupervised procedure over a large data set of celebrity faces. The auto-encoder latent space provides a meaningful, topologically organized 1024-dimensional description of each image. We then presented several thousand faces to human subjects, and learned a simple linear mapping between the multi-voxel fMRI activation patterns and the 1024 latent dimensions. Finally, we applied this mapping to novel test images, translating fMRI patterns into VAE latent codes, and codes into face reconstructions. The system not only performed robust pairwise decoding (>95% correct), but also accurate gender classification, and even decoded which face was imagined, rather than seen.

Traffic Flow Imputation Using Parallel Data and Generative Adversarial Networks

Traffic data imputation is critical for both research and applications of intelligent transportation systems. To develop traffic data imputation models with high accuracy, traffic data must be large and diverse, which is costly. An alternative is to use synthetic traffic data, which is cheap and easy-access. In this paper, we propose a novel approach using parallel data and generative adversarial networks (GANs) to enhance traffic data imputation. Parallel data is a recently proposed method of using synthetic and real data for data mining and data-driven process, in which we apply GANs to generate synthetic traffic data. As it is difficult for the standard GAN algorithm to generate time-dependent traffic flow data, we made twofold modifications: 1) using the real data or the corrupted ones instead of random vectors as latent codes to generator within GANs and 2) introducing a representation loss to measure discrepancy between the synthetic data and the real data. The experimental results on a real traffic dataset demonstrate that our method can significantly improve the performance of traffic data imputation.

Multimodal supervised image translation

Multimodal image-to-image translation is a class of vision and graphics problems where the goal is to learn a one-to-many mapping between the source domain and target domain. Given an image in the source domain, the model aims to produce as many diverse results as possible. It is an important and challenging problem in the task of image translation. To this end, recent works utilise Gaussian vectors to produce diverse results but with a small difference. It is because of the special probabilistic nature of Gaussian distribution. In this work, the authors propose linearly distributed latent codes instead of conventional Gaussian vectors, which control the style of generated images. Taking advantage of linear distribution, their model can produce much more diverse results and outperform the state-of-the-art baselines in terms of diversity. Qualitative and quantitative comparisons against baselines demonstrate the effectiveness and superiority of their method.

Deep learning cardiac motion analysis for human survival prediction.

Motion analysis is used in computer vision to understand the behaviour of moving objects in sequences of images. Optimising the interpretation of dynamic biological systems requires accurate and precise motion tracking as well as efficient representations of high-dimensional motion trajectories so that these can be used for prediction tasks. Here we use image sequences of the heart, acquired using cardiac magnetic resonance imaging, to create time-resolved three-dimensional segmentations using a fully convolutional network trained on anatomical shape priors. This dense motion model formed the input to a supervised denoising autoencoder (4Dsurvival), which is a hybrid network consisting of an autoencoder that learns a task-specific latent code representation trained on observed outcome data, yielding a latent representation optimised for survival prediction. To handle right-censored survival outcomes, our network used a Cox partial likelihood loss function. In a study of 302 patients the predictive accuracy (quantified by Harrell’s C-index) was significantly higher (p = .0012) for our model C=0.75 (95% CI: 0.70 - 0.79) than the human benchmark of C=0.59 (95% CI: 0.53 - 0.65). This work demonstrates how a complex computer vision task using high-dimensional medical image data can efficiently predict human survival.

A Novel and Efficient CVAE-GAN-Based Approach With Informative Manifold for Semi-Supervised Anomaly Detection

Semi-supervised anomaly detection identifies abnormal (testing) observations which are different from normal (training) observations. In many practical situations, anomalies are poorly insufficient and not well defined, while the normal data are easily sampled, have a wide variety, and may not be classified. For this paradigm, we propose a novel end-to-end deep network as an anomaly detector only trained on normal samples. Our architecture consists of a conditional variational auto-encoder (CVAE), a feature discriminator (FD), and an adversarially trained WGAN-GP discriminator. The CVAE is designed as a generator to reconstruct images. It leverages underlying category information and multivariate Gaussian distributions to regularize the latent space, enabling a smooth and informative manifold. For anomalies which have a certain similarity to normal data, we perform active negative training by generating potential outliers from the latent space to limit network generative capability. In order to capture data characteristics, we maximize the mutual information between the inputs and the latent codes by the FD. It enhances the relationship between the high-dimensional image space and corresponding encoded vectors. To promote reconstruction, a structural similarity loss is applied to robustly recover local texture details and the WGAN-GP discriminator is employed to aid in generating photo-realistic images. We distinguish anomalies by computing a reconstruction-based anomaly score. Different from recent encoder-decoder or GAN-based architectures, our approach considers input categories, constructs, and exploits a useful manifold in an unsupervised manner and has a stronger reconstruction capability. The experimental results demonstrate that the proposed approach outperforms state-of-the-art methods over several benchmark datasets.