Latent Code Research Articles

Radio Galaxy Zoo: Leveraging latent space representations from variational autoencoder

We propose to learn latent space representations of radio galaxies, and train a very deep variational autoencoder (VDVAE) on RGZ DR1, an unlabeled dataset, to this end. We show that the encoded features can be leveraged for downstream tasks such as classifying galaxies in labeled datasets, and similarity search. Results show that the model is able to reconstruct its given inputs, capturing the salient features of the latter. We use the latent codes of galaxy images, from MiraBest Confident and FR-DEEP NVSS datasets, to train various non-neural network classifiers. It is found that the latter can differentiate FRI from FRII galaxies achieving accuracy ≥ 76%, roc-auc ≥ 0.86, specificity ≥ 0.73 and recall ≥ 0.78 on MiraBest Confident dataset, comparable to results obtained in previous studies. The performance of simple classifiers trained on FR-DEEP NVSS data representations is on par with that of a deep learning classifier (CNN based) trained on images in previous work, highlighting how powerful the compressed information is. We successfully exploit the learned representations to search for galaxies in a dataset that are semantically similar to a query image belonging to a different dataset. Although generating new galaxy images (e.g. for data augmentation) is not our primary objective, we find that the VDVAE model is a relatively good emulator. Finally, as a step toward detecting anomaly/novelty, a density estimator — Masked Autoregressive Flow (MAF) — is trained on the latent codes, such that the log-likelihood of data can be estimated. The downstream tasks conducted in this work demonstrate the meaningfulness of the latent codes.

Application of C-InGAN Model in Interpretable Feature of Bearing Fault Diagnosis.

Although traditional fault diagnosis methods are proficient in extracting signal features, their diagnostic interpretability remains challenging. Consequently, this article proposes a conditionally interpretable generative adversarial network (C-InGAN) model for the interpretable feature fault diagnosis of bearings. Initially, the vibration signal is denoised and transformed into a frequency domain signal. The model consists of the two primary networks, each employing a convolutional layer and an attention module, generator (G) and discriminator (D), respectively. Latent code was incorporated into G to constrain the generated samples, and a discriminant layer was added to D to identify the interpretable features. During training, the two networks were alternately trained, and the feature mapping relationship of the pre-normalized encoder was learned by maximizing the information from the latent code and the discriminative result. The encoding that represents specific features in the vibration signal was extracted from the random noise. Ultimately, after completing adversarial learning, G is capable of generating a simulated signal of the specified feature, and D can assess the interpretable features in the vibration signal. The effectiveness of the model is validated through three typical experimental cases. This method effectively separates the discrete and continuous feature coding in the signal.

In-Domain GAN Inversion for Faithful Reconstruction and Editability.

Generative Adversarial Networks (GANs) have significantly advanced image synthesis through mapping randomly sampled latent codes to high-fidelity synthesized images. However, applying well-trained GANs to real image editing remains challenging. A common solution is to find an approximate latent code that can adequately recover the input image to edit, which is also known as GAN inversion. To invert a GAN model, prior works typically focus on reconstructing the target image at the pixel level, yet few studies are conducted on whether the inverted result can well support manipulation at the semantic level. This work fills in this gap by proposing in-domain GAN inversion, which consists of a domain-guided encoder and a domain-regularized optimizer, to regularize the inverted code in the native latent space of the pre-trained GAN model. In this way, we manage to sufficiently reuse the knowledge learned by GANs for image reconstruction, facilitating a wide range of editing applications without any retraining. We further make comprehensive analyses on the effects of the encoder structure, the starting inversion point, as well as the inversion parameter space, and observe the trade-off between the reconstruction quality and the editing property. Such a trade-off sheds light on how a GAN model represents an image with various semantics encoded in the learned latent distribution.

GenCoder: A Novel Convolutional Neural Network Based Autoencoder for Genomic Sequence Data Compression.

Revolutionary advances in DNA sequencing technologies fundamentally change the nature of genomics. Today's sequencing technologies have opened into an outburst in genomic data volume. These data can be used in various applications where long-term storage and analysis of genomic sequence data are required. Data-specific compression algorithms can effectively manage a large volume of data. In recent times, deep learning has achieved great success in many compression tools and is gradually being used in genomic sequence compression. Significantly, autoencoder has been applied in dimensionality reduction, compact representations of data, and generative model learning. It can use convolutional layers to learn essential features from input data, which is better for image and series data. Autoencoder reconstructs the input data with some loss of information. Since accuracy is critical in genomic data, compressed genomic data must be decompressed without any information loss. We introduce a new scheme to address the loss incurred in the decompressed data of the autoencoder. This paper proposes a novel algorithm called GenCoder for reference-free compression of genomic sequences using a convolutional autoencoder and regenerating the genomic sequences from a latent code produced by the autoencoder, and retrieving original data losslessly. Performance evaluation is conducted on various genomes and benchmarked datasets. The experimental results on the tested data demonstrate that the deep learning model used in the proposed compression algorithm generalizes well for genomic sequence data and achieves a compression gain of 27% over the best state-of-the-art method.

High-density generation of spatial transcriptomics with STAGE.

Spatial transcriptome technologies have enabled the measurement of gene expression while maintaining spatial location information for deciphering the spatial heterogeneity of biological tissues. However, they were heavily limited by the sparse spatial resolution and low data quality. To this end, we develop a spatial location-supervised auto-encoder generator STAGE for generating high-density spatial transcriptomics (ST). STAGE takes advantage of the customized supervised auto-encoder to learn continuous patterns of gene expression in space and generate high-resolution expressions for given spatial coordinates. STAGE can improve the low quality of spatial transcriptome data and smooth the generated manifold of gene expression through the de-noising function on the latent codes of the auto-encoder. Applications to four ST datasets, STAGE has shown better recovery performance for down-sampled data than existing methods, revealed significant tissue structure specificity, and enabled robust identification of spatially informative genes and patterns. In addition, STAGE can be extended to three-dimensional (3D) stacked ST data for generating gene expression at any position between consecutive sections for shaping high-density 3D ST configuration.

StylePart: image-based shape part manipulation

AbstractDirect part-level manipulation of man-made shapes in an image is desired given its simplicity. However, it is not intuitive given the existing manually created cuboid and cylinder controllers. To tackle this problem, we present StylePart, a framework that enables direct shape manipulation of an image by leveraging generative models of both images and 3D shapes. Our key contribution is a shape-consistent latent mapping function that connects the image generative latent space and the 3D man-made shape attribute latent space. Our method “forwardly maps” the image content to its corresponding 3D shape attributes, where the shape part can be easily manipulated. The attribute codes of the manipulated 3D shape are then “backwardly mapped” to the image latent code to obtain the final manipulated image. By using both forward and backward mapping, an user can edit the image directly without resorting to any 3D workflow. We demonstrate our approach through various manipulation tasks, including part replacement, part resizing, and shape orientation manipulation, and evaluate its effectiveness through extensive ablation studies.

Fast fluid–structure interaction simulation method based on deep learning flow field modeling

The rapid acquisition of high-fidelity flow field information is of great significance for engineering applications such as multi-field coupling. Current research in flow field modeling predominantly focuses on low Reynolds numbers and periodic flows, exhibiting weak generalization capabilities and notable issues with temporal inferring error accumulation. Therefore, we establish a reduced order model (ROM) based on Convolutional Auto-Encoder (CAE) and Long Short-Term Memory (LSTM) neural network and propose an unsteady flow field modeling method for the airfoil with a high Reynolds number and strong nonlinear characteristics. The attention mechanism and weak physical constraints are integrated into the model architecture to improve the modeling accuracy. A broadband excitation training strategy is proposed to overcome the error accumulation problem of long-term inferring. With only a small amount of latent codes, the relative error of the flow field reconstructed by CAE can be less than 5‰. By training LSTM with broadband excitation signals, stable dynamic evolution can be achieved in the time dimension. CAE-LSTM can accurately predict the forced response and complex limit cycle behavior of the airfoil in a wide range of amplitude and frequency under subsonic/transonic conditions. The relative errors of predicted variables and integral force are less than 1%. The fluid–structure interaction framework is built by coupling the ROM and motion equations of the structure. CAE-LSTM predicts the time series response of pitch displacement and moment coefficient at different reduced frequencies, which is in good agreement with computational fluid dynamics, and the simulation time savings exceed one order of magnitude.

Three Dimensional Microwave Data Inversion in Feature Space for Stroke Imaging.

Microwave imaging is a promising method for early diagnosing and monitoring brain strokes. It is portable, non-invasive, and safe to the human body. Conventional techniques solve for unknown electrical properties represented as pixels or voxels, but often result in inadequate structural information and high computational costs. We propose to reconstruct the three dimensional (3D) electrical properties of the human brain in a feature space, where the unknowns are latent codes of a variational autoencoder (VAE). The decoder of the VAE, with prior knowledge of the brain, acts as a module of data inversion. The codes in the feature space are optimized by minimizing the misfit between measured and simulated data. A dataset of 3D heads characterized by permittivity and conductivity is constructed to train the VAE. Numerical examples show that our method increases structural similarity by 14% and speeds up the solution process by over 3 orders of magnitude using only 4.8% number of the unknowns compared to the voxel-based method. This high-resolution imaging of electrical properties leads to more accurate stroke diagnosis and offers new insights into the study of the human brain.

Adversarial attacks on GAN-based image fusion

Image fusion has achieved significant success, owing to the rapid development of digital computing and Generative Adversarial Networks (GANs). GAN-based fusion techniques fuse latent codes through spatial or arithmetic operations to achieve real image fusion, facilitated by encoders. However, security concerns have arisen due to the vulnerability of deep neural networks to adversarial perturbations. This vulnerability extends to the GAN-based image fusion task. In this paper, we introduce two methods for creating adversarial examples in the context of GAN-based image fusion: adding subtle perturbations to input images or applying the adversarial patch to the input images. The subtle perturbation is meticulously crafted to be imperceptible yet capable of misleading a specific output, regardless of the input images. Conversely, the adversarial patch is a universal perturbation that, when applied to input images as a patch, induces meaningless output images. Our comprehensive experiments, conducted on datasets such as FFHQ (3 × 1024 × 1024) and Stanford Cars (3 × 512 × 512), include both quality and quantity evaluations. According to our experimental results, the subtle perturbation results in almost identical output images, while the adversarial patch induces meaningless fused images and can be transferred to other datasets. By demonstrating that image fusion models are highly vulnerable to adversarial attacks, this study highlights serious concerns regarding the security of these models.

Spatial-Contextual Discrepancy Information Compensation for GAN Inversion

Most existing GAN inversion methods either achieve accurate reconstruction but lack editability or offer strong editability at the cost of fidelity. Hence, how to balance the distortion-editability trade-off is a significant challenge for GAN inversion. To address this challenge, we introduce a novel spatial-contextual discrepancy information compensation-based GAN-inversion method (SDIC), which consists of a discrepancy information prediction network (DIPN) and a discrepancy information compensation network (DICN). SDIC follows a ``compensate-and-edit'' paradigm and successfully bridges the gap in image details between the original image and the reconstructed/edited image. On the one hand, DIPN encodes the multi-level spatial-contextual information of the original and initial reconstructed images and then predicts a spatial-contextual guided discrepancy map with two hourglass modules. In this way, a reliable discrepancy map that models the contextual relationship and captures fine-grained image details is learned. On the other hand, DICN incorporates the predicted discrepancy information into both the latent code and the GAN generator with different transformations, generating high-quality reconstructed/edited images. This effectively compensates for the loss of image details during GAN inversion. Both quantitative and qualitative experiments demonstrate that our proposed method achieves the excellent distortion-editability trade-off at a fast inference speed for both image inversion and editing tasks. Our code is available at https://github.com/ZzqLKED/SDIC.

Any-Stereo: Arbitrary Scale Disparity Estimation for Iterative Stereo Matching

Due to unaffordable computational costs, the regularized disparity in iterative stereo matching is typically maintained at a lower resolution than the input. To regress the full resolution disparity, most stereo methods resort to convolutions to decode a fixed-scale output. However, they are inadequate for recovering vital high-frequency information lost during downsampling, limiting their performance on full-resolution prediction. In this paper, we introduce AnyStereo, an accurate and efficient disparity upsampling module with implicit neural representation for the iterative stereo pipeline. By modeling the disparity as a continuous representation over 2D spatial coordinates, subtle details can emerge from the latent space at arbitrary resolution. To further complement the missing information and details in the latent code, we propose two strategies: intra-scale similarity unfolding and cross-scale feature alignment. The former unfolds the neighbor relationships, while the latter introduces the context in high-resolution feature maps. The proposed AnyStereo can seamlessly replace the upsampling module in most iterative stereo models, improving their ability to capture fine details and generate arbitrary-scale disparities even with fewer parameters. With our method, the iterative stereo pipeline establishes a new state-of-the-art performance. The code is available at https://github.com/Zhaohuai-L/Any-Stereo.

HyperEditor: Achieving Both Authenticity and Cross-Domain Capability in Image Editing via Hypernetworks

Editing real images authentically while also achieving cross-domain editing remains a challenge. Recent studies have focused on converting real images into latent codes and accomplishing image editing by manipulating these codes. However, merely manipulating the latent codes would constrain the edited images to the generator's image domain, hindering the attainment of diverse editing goals. In response, we propose an innovative image editing method called HyperEditor, which utilizes weight factors generated by hypernetworks to reassign the weights of the pre-trained StyleGAN2's generator. Guided by CLIP's cross-modal image-text semantic alignment, this innovative approach enables us to simultaneously accomplish authentic attribute editing and cross-domain style transfer, a capability not realized in previous methods. Additionally, we ascertain that modifying only the weights of specific layers in the generator can yield an equivalent editing result. Therefore, we introduce an adaptive layer selector, enabling our hypernetworks to autonomously identify the layers requiring output weight factors, which can further improve our hypernetworks' efficiency. Extensive experiments on abundant challenging datasets demonstrate the effectiveness of our method.

Gradual Residuals Alignment: A Dual-Stream Framework for GAN Inversion and Image Attribute Editing

GAN-based image attribute editing firstly leverages GAN Inversion to project real images into the latent space of GAN and then manipulates corresponding latent codes. Recent inversion methods mainly utilize additional high-bit features to improve image details preservation, as low-bit codes cannot faithfully reconstruct source images, leading to the loss of details. However, during editing, existing works fail to accurately complement the lost details and suffer from poor editability. The main reason is they inject all the lost details indiscriminately at one time, which inherently induces the position and quantity of details to overfit source images, resulting in inconsistent content and artifacts in edited images. This work argues that details should be gradually injected into both the reconstruction and editing process in a multi-stage coarse-to-fine manner for better detail preservation and high editability. Therefore, a novel dual-stream framework is proposed to accurately complement details at each stage. The Reconstruction Stream is employed to embed coarse-to-fine lost details into residual features and then adaptively add them to the GAN generator. In the Editing Stream, residual features are accurately aligned by our Selective Attention mechanism and then injected into the editing process in a multi-stage manner. Extensive experiments have shown the superiority of our framework in both reconstruction accuracy and editing quality compared with existing methods.

Collaborative Tooth Motion Diffusion Model in Digital Orthodontics

Tooth motion generation is an essential task in digital orthodontic treatment for precise and quick dental healthcare, which aims to generate the whole intermediate tooth motion process given the initial pathological and target ideal tooth alignments. Most prior works for multi-agent motion planning problems usually result in complex solutions. Moreover, the occlusal relationship between upper and lower teeth is often overlooked. In this paper, we propose a collaborative tooth motion diffusion model. The critical insight is to remodel the problem as a diffusion process. In this sense, we model the whole tooth motion distribution with a diffusion model and transform the planning problem into a sampling process from this distribution. We design a tooth latent representation to provide accurate conditional guides consisting of two key components: the tooth frame represents the position and posture, and the tooth latent shape code represents the geometric morphology. Subsequently, we present a collaborative diffusion model to learn the multi-tooth motion distribution based on inter-tooth and occlusal constraints, which are implemented by graph structure and new loss functions, respectively. Extensive qualitative and quantitative experiments demonstrate the superiority of our framework in the application of orthodontics compared with state-of-the-art methods.

G2L-CariGAN: Caricature Generation from Global Structure to Local Features

Existing GAN-based approaches to caricature generation mainly focus on exaggerating a character’s global facial structure. This often leads to the failure in highlighting significant facial features such as big eyes and hook nose. To address this limitation, we propose a new approach termed as G2L-CariGAN, which uses feature maps of spatial dimensions instead of latent codes for geometric exaggeration. G2L-CariGAN first exaggerates the global facial structure of the character on a low-dimensional feature map and then exaggerates its local facial features on a high-dimensional feature map. Moreover, we develop a caricature identity loss function based on feature maps, which well retains the character's identity after exaggeration. Our experiments have demonstrated that G2L-CariGAN outperforms the state-of-arts in terms of the quality of exaggerating a character and retaining its identity.

Constraint Latent Space Matters: An Anti-anomalous Waveform Transformation Solution from Photoplethysmography to Arterial Blood Pressure

Arterial blood pressure (ABP) holds substantial promise for proactive cardiovascular health management. Notwithstanding its potential, the invasive nature of ABP measurements confines their utility primarily to clinical environments, limiting their applicability for continuous monitoring beyond medical facilities. The conversion of photoplethysmography (PPG) signals into ABP equivalents has garnered significant attention due to its potential in revolutionizing cardiovascular disease management. Recent strides in PPG-to-ABP prediction encompass the integration of generative and discriminative models. Despite these advances, the efficacy of these models is curtailed by the latent space shift predicament, stemming from alterations in PPG data distribution across disparate hardware and individuals, potentially leading to distorted ABP waveforms. To tackle this problem, we present an innovative solution named the Latent Space Constraint Transformer (LSCT), leveraging a quantized codebook to yield robust latent spaces by employing multiple discretizing bases. To facilitate improved reconstruction, the Correlation-boosted Attention Module (CAM) is introduced to systematically query pertinent bases on a global scale. Furthermore, to enhance expressive capacity, we propose the Multi-Spectrum Enhancement Knowledge (MSEK), which fosters local information flow within the channels of latent code and provides additional embedding for reconstruction. Through comprehensive experimentation on both publicly available datasets and a private downstream task dataset, the proposed approach demonstrates noteworthy performance enhancements compared to existing methods. Extensive ablation studies further substantiate the effectiveness of each introduced module.

FaceRSA: RSA-Aware Facial Identity Cryptography Framework

With the flourishing of the Internet, sharing one's photos or automated processing of faces using computer vision technology has become an everyday occurrence. While enjoying the convenience, the concern for identity privacy is also emerging. Therefore, some efforts introduced the concept of ``password'' from traditional cryptography such as RSA into the face anonymization and deanonymization task to protect the facial identity without compromising the usability of the face image. However, these methods either suffer from the poor visual quality of the synthesis results or do not possess the full cryptographic properties, resulting in compromised security. In this paper, we present the first facial identity cryptography framework with full properties analogous to RSA. Our framework leverages the powerful generative capabilities of StyleGAN to achieve megapixel-level facial identity anonymization and deanonymization. Thanks to the great semantic decoupling of StyleGAN's latent space, the identity encryption and decryption process are performed in latent space by a well-designed password mapper in the manner of editing latent code. Meanwhile, the password-related information is imperceptibly hidden in the edited latent code owing to the redundant nature of the latent space. To make our cryptographic framework possesses all the properties analogous to RSA, we propose three types of loss functions: single anonymization loss, sequential anonymization loss, and associated anonymization loss. Extensive experiments and ablation analyses demonstrate the superiority of our method in terms of the quality of synthesis results, identity-irrelevant attributes preservation, deanonymization accuracy, and completeness of properties analogous to RSA.

WeditGAN: Few-Shot Image Generation via Latent Space Relocation

In few-shot image generation, directly training GAN models on just a handful of images faces the risk of overfitting. A popular solution is to transfer the models pretrained on large source domains to small target ones. In this work, we introduce WeditGAN, which realizes model transfer by editing the intermediate latent codes w in StyleGANs with learned constant offsets (delta w), discovering and constructing target latent spaces via simply relocating the distribution of source latent spaces. The established one-to-one mapping between latent spaces can naturally prevents mode collapse and overfitting. Besides, we also propose variants of WeditGAN to further enhance the relocation process by regularizing the direction or finetuning the intensity of delta w. Experiments on a collection of widely used source/target datasets manifest the capability of WeditGAN in generating realistic and diverse images, which is simple yet highly effective in the research area of few-shot image generation. Codes are available at https://github.com/Ldhlwh/WeditGAN.

Structured Probabilistic Coding

This paper presents a new supervised representation learning framework, namely structured probabilistic coding (SPC), to learn compact and informative representations from input related to the target task. SPC is an encoder-only probabilistic coding technology with a structured regularization from the target space. It can enhance the generalization ability of pre-trained language models for better language understanding. Specifically, our probabilistic coding simultaneously performs information encoding and task prediction in one module to more fully utilize the effective information from input data. It uses variational inference in the output space to reduce randomness and uncertainty. Besides, to better control the learning process of probabilistic representations, a structured regularization is proposed to promote uniformity across classes in the latent space. With the regularization term, SPC can preserve the Gaussian structure of the latent code and achieve better coverage of the hidden space with class uniformly. Experimental results on 12 natural language understanding tasks demonstrate that our SPC effectively improves the performance of pre-trained language models for classification and regression. Extensive experiments show that SPC can enhance the generalization capability, robustness to label noise, and clustering quality of output representations.

Implicit Modeling of Non-rigid Objects with Cross-Category Signals

Deep implicit functions (DIFs) have emerged as a potent and articulate means of representing 3D shapes. However, methods modeling object categories or non-rigid entities have mainly focused on single-object scenarios. In this work, we propose MODIF, a multi-object deep implicit function that jointly learns the deformation fields and instance-specific latent codes for multiple objects at once. Our emphasis is on non-rigid, non-interpenetrating entities such as organs. To effectively capture the interrelation between these entities and ensure precise, collision-free representations, our approach facilitates signaling between category-specific fields to adequately rectify shapes. We also introduce novel inter-object supervision: an attraction-repulsion loss is formulated to refine contact regions between objects. Our approach is demonstrated on various medical benchmarks, involving modeling different groups of intricate anatomical entities. Experimental results illustrate that our model can proficiently learn the shape representation of each organ and their relations to others, to the point that shapes missing from unseen instances can be consistently recovered by our method. Finally, MODIF can also propagate semantic information throughout the population via accurate point correspondences.