Recently, deep learning-based generative models have been introduced to generate singing voices. One approach is to predict the parametric vocoder features consisting of explicit speech parameters. This approach has the advantage that the meaning of each feature is explicitly distinguished. Another approach is to predict mel-spectrograms for a neural vocoder. However, parametric vocoders have limitations of voice quality and the mel-spectrogram features are difficult to model because the timbre and pitch information are entangled. In this study, we propose a singing voice synthesis model with multi-task learning to use both approaches -- acoustic features for a parametric vocoder and mel-spectrograms for a neural vocoder. By using the parametric vocoder features as auxiliary features, the proposed model can efficiently disentangle and control the timbre and pitch components of the mel-spectrogram. Moreover, a generative adversarial network framework is applied to improve the quality of singing voices in a multi-singer model. Experimental results demonstrate that our proposed model can generate more natural singing voices than the single-task models, while performing better than the conventional parametric vocoder-based model.
Deep learning (DL) techniques have been extensively utilized for medical image classification. Most DL-based classification networks are generally structured hierarchically and optimized through the minimization of a single loss function measured at the end of the networks. However, such a single loss design could potentially lead to optimization of one specific value of interest but fail to leverage informative features from intermediate layers that might benefit classification performance and reduce the risk of overfitting. Recently, auxiliary convolutional neural networks (AuxCNNs) have been employed on top of traditional classification networks to facilitate the training of intermediate layers to improve classification performance and robustness. In this study, we proposed an adversarial learning-based AuxCNN to support the training of deep neural networks for medical image classification. Two main innovations were adopted in our AuxCNN classification framework. First, the proposed AuxCNN architecture includes an image generator and an image discriminator for extracting more informative image features for medical image classification, motivated by the concept of generative adversarial network (GAN) and its impressive ability in approximating target data distribution. Second, a hybrid loss function is designed to guide the model training by incorporating different objectives of the classification network and AuxCNN to reduce overfitting. Comprehensive experimental studies demonstrated the superior classification performance of the proposed model. The effect of the network-related factors on classification performance was investigated.
Due to the pivotal role of Recommender Systems (RS) in guiding customers towards the purchase, there is a natural motivation for unscrupulous parties to spoof RS for profits. In this paper, we study Shilling Attack where an adversarial party injects a number of fake user profiles for improper purposes. Conventional Shilling Attack approaches lack attack transferability (i.e., attacks are not effective on some victim RS models) and/or attack invisibility (i.e., injected profiles can be easily detected). To overcome these issues, we present Leg-UP, a novel attack model based on the Generative Adversarial Network. Leg-UP learns user behavior patterns from real users in the sampled ``templates'' and constructs fake user profiles. To simulate real users, the generator in Leg-UP directly outputs discrete ratings. To enhance attack transferability, the parameters of the generator are optimized by maximizing the attack performance on a surrogate RS model. To improve attack invisibility, Leg-UP adopts a discriminator to guide the generator to generate undetectable fake user profiles. Experiments on benchmarks have shown that Leg-UP exceeds state-of-the-art Shilling Attack methods on a wide range of victim RS models. The source code of our work is available at: https://github.com/XMUDM/ShillingAttack.
We present a method for generating arrangements of indoor furniture from human-designed furniture layout data. Our method creates arrangements that target specified diversity, such as the total price of all furniture in the room and the number of pieces placed. To generate realistic furniture arrangement, we train a generative adversarial network (GAN) on human-designed layouts. To target specific diversity in the arrangements, we optimize the latent space of the GAN via a quality diversity algorithm to generate a diverse arrangement collection. Experiments show our approach discovers a set of arrangements that are similar to human-designed layouts but varies in price and number of furniture pieces.
This paper aims to quantify uncertainty in traffic state estimation (TSE) using the generative adversarial network based physics-informed deep learning (PIDL). The uncertainty of the focus arises from fundamental diagrams, in other words, the mapping from traffic density to velocity. To quantify uncertainty for the TSE problem is to characterize the robustness of predicted traffic states. Since its inception, generative adversarial networks (GAN) have become a popular probabilistic machine learning framework. In this paper, we will inform the GAN based predictions using stochastic traffic flow models and develop a GAN based PIDL framework for TSE, named ``PhysGAN-TSE". By conducting experiments on a real-world dataset, the Next Generation SIMulation (NGSIM) dataset, this method is shown to be more robust for uncertainty quantification than the pure GAN model or pure traffic flow models. Two physics models, the Lighthill-Whitham-Richards (LWR) and the Aw-Rascle-Zhang (ARZ) models, are compared as the physics components for the PhysGAN, and results show that the ARZ-based PhysGAN achieves a better performance than the LWR-based one.
This paper proposes the TrafficFlowGAN, a physics-informed flow based generative adversarial network (GAN), for uncertainty quantification (UQ) of dynamical systems. TrafficFlowGAN adopts a normalizing flow model as the generator to explicitly estimate the data likelihood. This flow model is trained to maximize the data likelihood and to generate synthetic data that can fool a convolutional discriminator. We further regularize this training process using prior physics information, so-called physics-informed deep learning (PIDL). To the best of our knowledge, we are the first to propose an integration of flow, GAN and PIDL for the UQ problems. We take the traffic state estimation (TSE), which aims to estimate the traffic variables (e.g. traffic density and velocity) using partially observed data, as an example to demonstrate the performance of our proposed model. We conduct numerical experiments where the proposed model is applied to learn the solutions of stochastic differential equations. The results demonstrate the robustness and accuracy of the proposed model, together with the ability to learn a machine learning surrogate model. We also test it on a real-world dataset, the Next Generation SIMulation (NGSIM), to show that the proposed TrafficFlowGAN can outperform the baselines, including the pure flow model, the physics-informed flow model, and the flow based GAN model.
Generative Adversarial Network (GAN) is one of the state-of-the-art generative models for realistic image synthesis. While training and evaluating GAN becomes increasingly important, the current GAN research ecosystem does not provide reliable benchmarks for which the evaluation is conducted consistently and fairly. Furthermore, because there are few validated GAN implementations, researchers devote considerable time to reproducing baselines. We study the taxonomy of GAN approaches and present a new open-source library named StudioGAN. StudioGAN supports 7 GAN architectures, 9 conditioning methods, 4 adversarial losses, 12 regularization modules, 3 differentiable augmentations, 7 evaluation metrics, and 5 evaluation backbones. With our training and evaluation protocol, we present a large-scale benchmark using various datasets (CIFAR10, ImageNet, AFHQv2, FFHQ, and Baby/Papa/Granpa-ImageNet) and 3 different evaluation backbones (InceptionV3, SwAV, and Swin Transformer). Unlike other benchmarks used in the GAN community, we train representative GANs, including BigGAN and StyleGAN series in a unified training pipeline and quantify generation performance with 7 evaluation metrics. The benchmark evaluates other cutting-edge generative models (e.g., StyleGAN-XL, ADM, MaskGIT, and RQ-Transformer). StudioGAN provides GAN implementations, training, and evaluation scripts with the pre-trained weights. StudioGAN is available at https://github.com/POSTECH-CVLab/PyTorch-StudioGAN.
Automatic deep learning segmentation models has been shown to improve both the segmentation efficiency and the accuracy. However, training a robust segmentation model requires considerably large labeled training samples, which may be impractical. This study aimed to develop a deep learning framework for generating synthetic lesions that can be used to enhance network training. The lesion synthesis network is a modified generative adversarial network (GAN). Specifically, we innovated a partial convolution strategy to construct an Unet-like generator. The discriminator is designed using Wasserstein GAN with gradient penalty and spectral normalization. A mask generation method based on principal component analysis was developed to model various lesion shapes. The generated masks are then converted into liver lesions through a lesion synthesis network. The lesion synthesis framework was evaluated for lesion textures, and the synthetic lesions were used to train a lesion segmentation network to further validate the effectiveness of this framework. All the networks are trained and tested on the public dataset from LITS. The synthetic lesions generated by the proposed approach have very similar histogram distributions compared to the real lesions for the two employed texture parameters, GLCM-energy and GLCM-correlation. The Kullback-Leibler divergence of GLCM-energy and GLCM-correlation were 0.01 and 0.10, respectively. Including the synthetic lesions in the tumor segmentation network improved the segmentation dice performance of U-Net significantly from 67.3% to 71.4% (p<0.05). Meanwhile, the volume precision and sensitivity improve from 74.6% to 76.0% (p=0.23) and 66.1% to 70.9% (p<0.01), respectively. The synthetic data significantly improves the segmentation performance.
Recent deep face recognition models proposed in the literature utilized large-scale public datasets such as MS-Celeb-1M and VGGFace2 for training very deep neural networks, achieving state-of-the-art performance on mainstream benchmarks. Recently, many of these datasets, e.g., MS-Celeb-1M and VGGFace2, are retracted due to credible privacy and ethical concerns. This motivates this work to propose and investigate the feasibility of using a privacy-friendly synthetically generated face dataset to train face recognition models. Towards this end, we utilize a class-conditional generative adversarial network to generate class-labeled synthetic face images, namely SFace. To address the privacy aspect of using such data to train a face recognition model, we provide extensive evaluation experiments on the identity relation between the synthetic dataset and the original authentic dataset used to train the generative model. Our reported evaluation proved that associating an identity of the authentic dataset to one with the same class label in the synthetic dataset is hardly possible. We also propose to train face recognition on our privacy-friendly dataset, SFace, using three different learning strategies, multi-class classification, label-free knowledge transfer, and combined learning of multi-class classification and knowledge transfer. The reported evaluation results on five authentic face benchmarks demonstrated that the privacy-friendly synthetic dataset has high potential to be used for training face recognition models, achieving, for example, a verification accuracy of 91.87\% on LFW using multi-class classification and 99.13\% using the combined learning strategy.
Many robotic tasks are composed of a lot of temporally correlated sub-tasks in a highly complex environment. It is important to discover situational intentions and proper actions by deliberating on temporal abstractions to solve problems effectively. To understand the intention separated from changing task dynamics, we extend an empowerment-based regularization technique to situations with multiple tasks based on the framework of a generative adversarial network. Under the multitask environments with unknown dynamics, we focus on learning a reward and policy from the unlabeled expert examples. In this study, we define situational empowerment as the maximum of mutual information representing how an action conditioned on both a certain state and sub-task affects the future. Our proposed method derives the variational lower bound of the situational mutual information to optimize it. We simultaneously learn the transferable multi-task reward function and policy by adding an induced term to the objective function. By doing so, the multi-task reward function helps to learn a robust policy for environmental change. We validate the advantages of our approach on multi-task learning and multi-task transfer learning. We demonstrate our proposed method has the robustness of both randomness and changing task dynamics. Finally, we prove that our method has significantly better performance and data efficiency than existing imitation learning methods on various benchmarks.
Screening colonoscopy is an important clinical application for several 3D computer vision techniques, including depth estimation, surface reconstruction, and missing region detection. However, the development, evaluation, and comparison of these techniques in real colonoscopy videos remain largely qualitative due to the difficulty of acquiring ground truth data. In this work, we present a Colonoscopy 3D Video Dataset (C3VD) acquired with a high definition clinical colonoscope and high-fidelity colon models for benchmarking computer vision methods in colonoscopy. We introduce a novel multimodal 2D-3D registration technique to register optical video sequences with ground truth rendered views of a known 3D model. The different modalities are registered by transforming optical images to depth maps with a Generative Adversarial Network and aligning edge features with an evolutionary optimizer. This registration method achieves an average translation error of 0.321 millimeters and an average rotation error of 0.159 degrees in simulation experiments where error-free ground truth is available. The method also leverages video information, improving registration accuracy by 55.6% for translation and 60.4% for rotation compared to single frame registration. 22 short video sequences were registered to generate 10,015 total frames with paired ground truth depth, surface normals, optical flow, occlusion, six degree-of-freedom pose, coverage maps, and 3D models. The dataset also includes screening videos acquired by a gastroenterologist with paired ground truth pose and 3D surface models. The dataset and registration source code are available at durr.jhu.edu/C3VD.
Novel photo-realistic texture synthesis is an important task for generating novel scenes, including asset generation for 3D simulations. However, to date, these methods predominantly generate textured objects in 2D space. If we rely on 2D object generation, then we need to make a computationally expensive forward pass each time we change the camera viewpoint or lighting. Recent work that can generate textures in 3D requires 3D component segmentation that is expensive to acquire. In this work, we present a novel conditional generative architecture that we call a graph generative adversarial network (GGAN) that can generate textures in 3D by learning object component information in an unsupervised way. In this framework, we do not need an expensive forward pass whenever the camera viewpoint or lighting changes, and we do not need expensive 3D part information for training, yet the model can generalize to unseen 3D meshes and generate appropriate novel 3D textures. We compare this approach against state-of-the-art texture generation methods and demonstrate that the GGAN obtains significantly better texture generation quality (according to Frechet inception distance). We release our model source code as open source.
An optical microscopic examination of thinly cut stained tissue on glass slides prepared from a FFPE tissue blocks is the gold standard for tissue diagnostics. In addition, the diagnostic abilities and expertise of any pathologist is dependent on their direct experience with common as well as rarer variant morphologies. Recently, deep learning approaches have been used to successfully show a high level of accuracy for such tasks. However, obtaining expert-level annotated images is an expensive and time-consuming task and artificially synthesized histological images can prove greatly beneficial. Here, we present an approach to not only generate histological images that reproduce the diagnostic morphologic features of common disease but also provide a user ability to generate new and rare morphologies. Our approach involves developing a generative adversarial network model that synthesizes pathology images constrained by class labels. We investigated the ability of this framework in synthesizing realistic prostate and colon tissue images and assessed the utility of these images in augmenting diagnostic ability of machine learning methods as well as their usability by a panel of experienced anatomic pathologists. Synthetic data generated by our framework performed similar to real data in training a deep learning model for diagnosis. Pathologists were not able to distinguish between real and synthetic images and showed a similar level of inter-observer agreement for prostate cancer grading. We extended the approach to significantly more complex images from colon biopsies and showed that the complex microenvironment in such tissues can also be reproduced. Finally, we present the ability for a user to generate deepfake histological images via a simple markup of sematic labels.
Adversarial training (AT) is a simple yet effective defense against adversarial attacks to image classification systems, which is based on augmenting the training set with attacks that maximize the loss. However, the effectiveness of AT as a defense for video classification has not been thoroughly studied. Our first contribution is to show that generating optimal attacks for video requires carefully tuning the attack parameters, especially the step size. Notably, we show that the optimal step size varies linearly with the attack budget. Our second contribution is to show that using a smaller (sub-optimal) attack budget at training time leads to a more robust performance at test time. Based on these findings, we propose three defenses against attacks with variable attack budgets. The first one, Adaptive AT, is a technique where the attack budget is drawn from a distribution that is adapted as training iterations proceed. The second, Curriculum AT, is a technique where the attack budget is increased as training iterations proceed. The third, Generative AT, further couples AT with a denoising generative adversarial network to boost robust performance. Experiments on the UCF101 dataset demonstrate that the proposed methods improve adversarial robustness against multiple attack types.
In this paper we propose a generative adversarial network (GAN) framework to enhance the perceptual quality of compressed videos. Our framework includes attention and adaptation to different quantization parameters (QPs) in a single model. The attention module exploits global receptive fields that can capture and align long-range correlations between consecutive frames, which can be beneficial for enhancing perceptual quality of videos. The frame to be enhanced is fed into the deep network together with its neighboring frames, and in the first stage features at different depths are extracted. Then extracted features are fed into attention blocks to explore global temporal correlations, followed by a series of upsampling and convolution layers. Finally, the resulting features are processed by the QP-conditional adaptation module which leverages the corresponding QP information. In this way, a single model can be used to enhance adaptively to various QPs without requiring multiple models specific for every QP value, while having similar performance. Experimental results demonstrate the superior performance of the proposed PeQuENet compared with the state-of-the-art compressed video quality enhancement algorithms.
Prognostics aid in the longevity of fielded systems or products. Quantifying the system's current health enable prognosis to enhance the operator's decision-making to preserve the system's health. Creating a prognosis for a system can be difficult due to (a) unknown physical relationships and/or (b) irregularities in data appearing well beyond the initiation of a problem. Traditionally, three different modeling paradigms have been used to develop a prognostics model: physics-based (PbM), data-driven (DDM), and hybrid modeling. Recently, the hybrid modeling approach that combines the strength of both PbM and DDM based approaches and alleviates their limitations is gaining traction in the prognostics domain. In this paper, a novel hybrid modeling approach for prognostics applications based on combining concepts from fuzzy logic and generative adversarial networks (GANs) is outlined. The FuzzyGAN based method embeds a physics-based model in the aggregation of the fuzzy implications. This technique constrains the output of the learning method to a realistic solution. Results on a bearing problem showcases the efficacy of adding a physics-based aggregation in a fuzzy logic model to improve GAN's ability to model health and give a more accurate system prognosis.
While a multitude of studies have been conducted on graph drawing, many existing methods only focus on optimizing a single aesthetic aspect of graph layouts, which can lead to sub-optimal results. There are a few existing methods that have attempted to develop a flexible solution for optimizing different aesthetic aspects measured by different aesthetic criteria. Furthermore, thanks to the significant advance in deep learning techniques, several deep learning-based layout methods were proposed recently. These methods have demonstrated the advantages of deep learning approaches for graph drawing. However, none of these existing methods can be directly applied to optimizing non-differentiable criteria without special accommodation. In this work, we propose a novel Generative Adversarial Network (GAN) based deep learning framework for graph drawing, called SmartGD, which can optimize different quantitative aesthetic goals, regardless of their differentiability. To demonstrate the effectiveness and efficiency of SmartGD, we conducted experiments on minimizing stress, minimizing edge crossing, maximizing crossing angle, maximizing shape-based metrics, and a combination of multiple aesthetics. Compared with several popular graph drawing algorithms, the experimental results show that SmartGD achieves good performance both quantitatively and qualitatively.
Matching has become the mainstream in counterfactual inference, with which selection bias between sample groups can be significantly eliminated. However in practice, when estimating average treatment effect on the treated (ATT) via matching, no matter which method, the trade-off between estimation accuracy and information loss constantly exist. Attempting to completely replace the matching process, this paper proposes the GAN-ATT estimator that integrates generative adversarial network (GAN) into counterfactual inference framework. Through GAN machine learning, the probability density functions (PDFs) of samples in both treatment group and control group can be approximated. By differentiating conditional PDFs of the two groups with identical input condition, the conditional average treatment effect (CATE) can be estimated, and the ensemble average of corresponding CATEs over all treatment group samples is the estimate of ATT. Utilizing GAN-based infinite sample augmentations, problems in the case of insufficient samples or lack of common support domains can be easily solved. Theoretically, when GAN could perfectly learn the PDFs, our estimators can provide exact estimate of ATT. To check the performance of the GAN-ATT estimator, three sets of data are used for ATT estimations: Two toy data sets with 1/2 dimensional covariate inputs and constant/covariate-dependent treatment effect are tested. The estimates of GAN-ATT are proved close to the ground truth and are better than traditional matching approaches; A real firm-level data set with high-dimensional input is tested and the applicability towards real data sets is evaluated by comparing matching approaches. Through the evidences obtained from the three tests, we believe that the GAN-ATT estimator has significant advantages over traditional matching methods in estimating ATT.
Federated learning (FL) has recently emerged as a popular privacy-preserving collaborative learning paradigm. However, it suffers from the non-independent and identically distributed (non-IID) data among clients. In this paper, we propose a novel framework, named Synthetic Data Aided Federated Learning (SDA-FL), to resolve this non-IID challenge by sharing synthetic data. Specifically, each client pretrains a local generative adversarial network (GAN) to generate differentially private synthetic data, which are uploaded to the parameter server (PS) to construct a global shared synthetic dataset. To generate confident pseudo labels for the synthetic dataset, we also propose an iterative pseudo labeling mechanism performed by the PS. A combination of the local private dataset and synthetic dataset with confident pseudo labels leads to nearly identical data distributions among clients, which improves the consistency among local models and benefits the global aggregation. Extensive experiments evidence that the proposed framework outperforms the baseline methods by a large margin in several benchmark datasets under both the supervised and semi-supervised settings.
An ideal music synthesizer should be both interactive and expressive, generating high-fidelity audio in realtime for arbitrary combinations of instruments and notes. Recent neural synthesizers have exhibited a tradeoff between domain-specific models that offer detailed control of only specific instruments, or raw waveform models that can train on any music but with minimal control and slow generation. In this work, we focus on a middle ground of neural synthesizers that can generate audio from MIDI sequences with arbitrary combinations of instruments in realtime. This enables training on a wide range of transcription datasets with a single model, which in turn offers note-level control of composition and instrumentation across a wide range of instruments. We use a simple two-stage process: MIDI to spectrograms with an encoder-decoder Transformer, then spectrograms to audio with a generative adversarial network (GAN) spectrogram inverter. We compare training the decoder as an autoregressive model and as a Denoising Diffusion Probabilistic Model (DDPM) and find that the DDPM approach is superior both qualitatively and as measured by audio reconstruction and Fr\'echet distance metrics. Given the interactivity and generality of this approach, we find this to be a promising first step towards interactive and expressive neural synthesis for arbitrary combinations of instruments and notes.