Search

The Diffusion Transformers Models (DiTs) have transitioned the network architecture from traditional UNets to transformers, demonstrating exceptional capabilities in image generation. Although DiTs have been widely applied to high-definition video generation tasks, their large parameter size hinders inference on edge devices. Vector quantization (VQ) can decompose model weight into a codebook and assignments, allowing extreme weight quantization and significantly reducing memory usage. In this paper, we propose VQ4DiT, a fast post-training vector quantization method for DiTs. We found that traditional VQ methods calibrate only the codebook without calibrating the assignments. This leads to weight sub-vectors being incorrectly assigned to the same assignment, providing inconsistent gradients to the codebook and resulting in a suboptimal result. To address this challenge, VQ4DiT calculates the candidate assignment set for each weight sub-vector based on Euclidean distance and reconstructs the sub-vector based on the weighted average. Then, using the zero-data and block-wise calibration method, the optimal assignment from the set is efficiently selected while calibrating the codebook. VQ4DiT quantizes a DiT XL/2 model on a single NVIDIA A100 GPU within 20 minutes to 5 hours depending on the different quantization settings. Experiments show that VQ4DiT establishes a new state-of-the-art in model size and performance trade-offs, quantizing weights to 2-bit precision while retaining acceptable image generation quality., Comment: 11 pages, 6 figures

Multi-exposure High Dynamic Range (HDR) imaging is a challenging task when facing truncated texture and complex motion. Existing deep learning-based methods have achieved great success by either following the alignment and fusion pipeline or utilizing attention mechanism. However, the large computation cost and inference delay hinder them from deploying on resource limited devices. In this paper, to achieve better efficiency, a novel Selective Alignment Fusion Network (SAFNet) for HDR imaging is proposed. After extracting pyramid features, it jointly refines valuable area masks and cross-exposure motion in selected regions with shared decoders, and then fuses high quality HDR image in an explicit way. This approach can focus the model on finding valuable regions while estimating their easily detectable and meaningful motion. For further detail enhancement, a lightweight refine module is introduced which enjoys privileges from previous optical flow, selection masks and initial prediction. Moreover, to facilitate learning on samples with large motion, a new window partition cropping method is presented during training. Experiments on public and newly developed challenging datasets show that proposed SAFNet not only exceeds previous SOTA competitors quantitatively and qualitatively, but also runs order of magnitude faster. Code and dataset is available at https://github.com/ltkong218/SAFNet., Comment: Accepted by ECCV 2024

Self-supervised monocular depth estimation has gathered notable interest since it can liberate training from dependency on depth annotations. In monocular video training case, recent methods only conduct view synthesis between existing camera views, leading to insufficient guidance. To tackle this, we try to synthesize more virtual camera views by flow-based video frame interpolation (VFI), termed as temporal augmentation. For multi-frame inference, to sidestep the problem of dynamic objects encountered by explicit geometry-based methods like ManyDepth, we return to the feature fusion paradigm and design a VFI-assisted multi-frame fusion module to align and aggregate multi-frame features, using motion and occlusion information obtained by the flow-based VFI model. Finally, we construct a unified self-supervised learning framework, named Mono-ViFI, to bilaterally connect single- and multi-frame depth. In this framework, spatial data augmentation through image affine transformation is incorporated for data diversity, along with a triplet depth consistency loss for regularization. The single- and multi-frame models can share weights, making our framework compact and memory-efficient. Extensive experiments demonstrate that our method can bring significant improvements to current advanced architectures. Source code is available at https://github.com/LiuJF1226/Mono-ViFI., Comment: 27 pages, accepted by ECCV 2024

Although recent years have witnessed significant advancements in image editing thanks to the remarkable progress of text-to-image diffusion models, the problem of non-rigid image editing still presents its complexities and challenges. Existing methods often fail to achieve consistent results due to the absence of unique identity characteristics. Thus, learning a personalized identity prior might help with consistency in the edited results. In this paper, we explore a novel task: learning the personalized identity prior for text-based non-rigid image editing. To address the problems in jointly learning prior and editing the image, we present LIPE, a two-stage framework designed to customize the generative model utilizing a limited set of images of the same subject, and subsequently employ the model with learned prior for non-rigid image editing. Experimental results demonstrate the advantages of our approach in various editing scenarios over past related leading methods in qualitative and quantitative ways.

Efficient and precise calculations of thermal transport properties and figure of merit, alongside a deep comprehension of thermal transport mechanisms, are essential for the practical utilization of advanced thermoelectric materials. In this study, we explore the microscopic processes governing thermal transport in the distinguished crystalline material Tl$_9$SbTe$_6$ by integrating a unified thermal transport theory with machine learning-assisted self-consistent phonon calculations. Leveraging machine learning potentials, we expedite the analysis of phonon energy shifts, higher-order scattering mechanisms, and thermal conductivity arising from various contributing factors like population and coherence channels. Our finding unveils an exceptionally low thermal conductivity of 0.31 W m$^{-1}$ K$^{-1}$ at room temperature, a result that closely correlates with experimental observations. Notably, we observe that the off-diagonal terms of heat flux operators play a significant role in shaping the overall lattice thermal conductivity of Tl$_9$SbTe$_6$, where the ultralow thermal conductivity resembles that of glass due to limited group velocities. Furthermore, we achieve a maximum $ZT$ value of 3.17 in the $c$-axis orientation for \textit{p}-type Tl$_9$SbTe$_6$ at 600 K, and an optimal $ZT$ value of 2.26 in the $a$-axis and $b$-axis direction for \textit{n}-type Tl$_9$SbTe$_6$ at 500 K. The crystalline Tl$_9$SbTe$_6$ not only showcases remarkable thermal insulation but also demonstrates impressive electrical properties owing to the dual-degeneracy phenomenon within its valence band. These results not only elucidate the underlying reasons for the exceptional thermoelectric performance of Tl$_9$SbTe$_6$ but also suggest potential avenues for further experimental exploration.

Precise image editing with text-to-image models has attracted increasing interest due to their remarkable generative capabilities and user-friendly nature. However, such attempts face the pivotal challenge of misalignment between the intended precise editing target regions and the broader area impacted by the guidance in practice. Despite excellent methods leveraging attention mechanisms that have been developed to refine the editing guidance, these approaches necessitate modifications through complex network architecture and are limited to specific editing tasks. In this work, we re-examine the diffusion process and misalignment problem from a frequency perspective, revealing that, due to the power law of natural images and the decaying noise schedule, the denoising network primarily recovers low-frequency image components during the earlier timesteps and thus brings excessive low-frequency signals for editing. Leveraging this insight, we introduce a novel fine-tuning free approach that employs progressive $\textbf{Fre}$qu$\textbf{e}$ncy truncation to refine the guidance of $\textbf{Diff}$usion models for universal editing tasks ($\textbf{FreeDiff}$). Our method achieves comparable results with state-of-the-art methods across a variety of editing tasks and on a diverse set of images, highlighting its potential as a versatile tool in image editing applications., Comment: Accepted by ECCV-2024

In this work, we target the task of text-driven style transfer in the context of text-to-image (T2I) diffusion models. The main challenge is consistent structure preservation while enabling effective style transfer effects. The past approaches in this field directly concatenate the content and style prompts for a prompt-level style injection, leading to unavoidable structure distortions. In this work, we propose a novel solution to the text-driven style transfer task, namely, Adaptive Style Incorporation~(ASI), to achieve fine-grained feature-level style incorporation. It consists of the Siamese Cross-Attention~(SiCA) to decouple the single-track cross-attention to a dual-track structure to obtain separate content and style features, and the Adaptive Content-Style Blending (AdaBlending) module to couple the content and style information from a structure-consistent manner. Experimentally, our method exhibits much better performance in both structure preservation and stylized effects.

In recent years, instruction-based image editing methods have garnered significant attention in image editing. However, despite encompassing a wide range of editing priors, these methods are helpless when handling editing tasks that are challenging to accurately describe through language. We propose InstructBrush, an inversion method for instruction-based image editing methods to bridge this gap. It extracts editing effects from exemplar image pairs as editing instructions, which are further applied for image editing. Two key techniques are introduced into InstructBrush, Attention-based Instruction Optimization and Transformation-oriented Instruction Initialization, to address the limitations of the previous method in terms of inversion effects and instruction generalization. To explore the ability of instruction inversion methods to guide image editing in open scenarios, we establish a TransformationOriented Paired Benchmark (TOP-Bench), which contains a rich set of scenes and editing types. The creation of this benchmark paves the way for further exploration of instruction inversion. Quantitatively and qualitatively, our approach achieves superior performance in editing and is more semantically consistent with the target editing effects., Comment: Project Page: https://royzhao926.github.io/InstructBrush/

Facial Appearance Editing (FAE) aims to modify physical attributes, such as pose, expression and lighting, of human facial images while preserving attributes like identity and background, showing great importance in photograph. In spite of the great progress in this area, current researches generally meet three challenges: low generation fidelity, poor attribute preservation, and inefficient inference. To overcome above challenges, this paper presents DiffFAE, a one-stage and highly-efficient diffusion-based framework tailored for high-fidelity FAE. For high-fidelity query attributes transfer, we adopt Space-sensitive Physical Customization (SPC), which ensures the fidelity and generalization ability by utilizing rendering texture derived from 3D Morphable Model (3DMM). In order to preserve source attributes, we introduce the Region-responsive Semantic Composition (RSC). This module is guided to learn decoupled source-regarding features, thereby better preserving the identity and alleviating artifacts from non-facial attributes such as hair, clothes, and background. We further introduce a consistency regularization for our pipeline to enhance editing controllability by leveraging prior knowledge in the attention matrices of diffusion model. Extensive experiments demonstrate the superiority of DiffFAE over existing methods, achieving state-of-the-art performance in facial appearance editing.

Maxwell's equations are a collection of coupled partial differential equations (PDEs) that, together with the Lorentz force law, constitute the basis of classical electromagnetism and electric circuits. Effectively solving Maxwell's equations is crucial in various fields, like electromagnetic scattering and antenna design optimization. Physics-informed neural networks (PINNs) have shown powerful ability in solving PDEs. However, PINNs still struggle to solve Maxwell's equations in heterogeneous media. To this end, we propose a domain-adaptive PINN (da-PINN) to solve inverse problems of Maxwell's equations in heterogeneous media. First, we propose a location parameter of media interface to decompose the whole domain into several sub-domains. Furthermore, the electromagnetic interface conditions are incorporated into a loss function to improve the prediction performance near the interface. Then, we propose a domain-adaptive training strategy for da-PINN. Finally, the effectiveness of da-PINN is verified with two case studies., Comment: 5 pages,4 figures

We present precise photometric estimates of stellar parameters, including effective temperature, metallicity, luminosity classification, distance, and stellar age, for nearly 26 million stars using the methodology developed in the first paper of this series, based on the stellar colors from the Stellar Abundances and Galactic Evolution Survey (SAGES) DR1 and Gaia EDR3. The optimal design of stellar-parameter sensitive $uv$ filters by SAGES has enabled us to determine photometric-metallicity estimates down to $-3.5$, similar to our previous results with the SkyMapper Southern Survey (SMSS), yielding a large sample of over five million metal-poor (MP; [Fe/H]$\le -1.0$) stars and nearly one million very metal-poor (VMP; [Fe/H]$\le -2.0$) stars. The typical precision is around $0.1$ dex for both dwarf and giant stars with [Fe/H]$>-1.0$, and 0.15-0.25/0.3-0.4 dex for dwarf/giant stars with [Fe/H]$<-1.0$. Using the precise parallax measurements and stellar colors from Gaia, effective temperature, luminosity classification, distance and stellar age are further derived for our sample stars. This huge data set in the Northern sky from SAGES, together with similar data in the Southern sky from SMSS, will greatly advance our understanding of the Milky Way, in particular its formation and evolution., Comment: 14 pages, 14 figures, 3 tables, accepted by ApJ. arXiv admin note: text overlap with arXiv:2104.14154

Adversarial examples bring a considerable security threat to support vector machines (SVMs), especially those used in safety-critical applications. Thus, robustness verification is an essential issue for SVMs, which can provide provable robustness against various kinds of adversary attacks. The evaluation results obtained through the robustness verification can provide a safe guarantee for the use of SVMs. The existing verification method does not often perform well in verifying SVMs with nonlinear kernels. To this end, we propose a method to improve the verification performance for SVMs with nonlinear kernels. We first formalize the adversarial robustness evaluation of SVMs as an optimization problem. Then a lower bound of the original problem is obtained by solving the Lagrangian dual problem of the original problem. Finally, the adversarial robustness of SVMs is evaluated concerning the lower bound. We evaluate the adversarial robustness of SVMs with linear and nonlinear kernels on the MNIST and Fashion-MNIST datasets. The experimental results show that the percentage of provable robustness obtained by our method on the test set is better than that of the state-of-the-art.

By creating networks of biochemical pathways, communities of micro-organisms are able to modulate the properties of their environment and even the metabolic processes within their hosts. Next-generation high-throughput sequencing has led to a new frontier in microbial ecology, promising the ability to leverage the microbiome to make crucial advancements in the environmental and biomedical sciences. However, this is challenging, as genomic data are high-dimensional, sparse, and noisy. Much of this noise reflects the exact conditions under which sequencing took place, and is so significant that it limits consensus-based validation of study results. We propose an ensemble approach for cross-study exploratory analyses of microbial abundance data in which we first estimate the variance-covariance matrix of the underlying abundances from each dataset on the log scale assuming Poisson sampling, and subsequently model these covariances jointly so as to find a shared low-dimensional subspace of the feature space. By viewing the projection of the latent true abundances onto this common structure, the variation is pared down to that which is shared among all datasets, and is likely to reflect more generalizable biological signal than can be inferred from individual datasets. We investigate several ways of achieving this, and demonstrate that they work well on simulated and real metagenomic data in terms of signal retention and interpretability.

Accurate mapping of forests is critical for forest management and carbon stocks monitoring. Deep learning is becoming more popular in Earth Observation (EO), however, the availability of reference data limits its potential in wide-area forest mapping. To overcome those limitations, here we introduce contrastive regression into EO based forest mapping and develop a novel semisupervised regression framework for wall-to-wall mapping of continuous forest variables. It combines supervised contrastive regression loss and semi-supervised Cross-Pseudo Regression loss. The framework is demonstrated over a boreal forest site using Copernicus Sentinel-1 and Sentinel-2 imagery for mapping forest tree height. Achieved prediction accuracies are strongly better compared to using vanilla UNet or traditional regression models, with relative RMSE of 15.1% on stand level. We expect that developed framework can be used for modeling other forest variables and EO datasets.

Non-Negative Matrix Factorization (NMF) is a widely used dimension reduction method that factorizes a non-negative data matrix into two lower dimensional non-negative matrices: One is the basis or feature matrix which consists of the variables and the other is the coefficients matrix which is the projections of data points to the new basis. The features can be interpreted as sub-structures of the data. The number of sub-structures in the feature matrix is also called the rank which is the only tuning parameter in NMF. An appropriate rank will extract the key latent features while minimizing the noise from the original data. In this paper, we develop a novel rank selection method based on hypothesis testing, using a deconvolved bootstrap distribution to assess the significance level accurately despite the large amount of optimization error. In the simulation section, we compare our method with a rank selection method based on hypothesis testing using bootstrap distribution without deconvolution, and with a cross-validated imputation method1. Through simulations, we demonstrate that our method is not only accurate at estimating the true ranks for NMF especially when the features are hard to distinguish but also efficient at computation. When applied to real microbiome data (e.g. OTU data and functional metagenomic data), our method also shows the ability to extract interpretable sub-communities in the data.

Deconvolution is the important problem of estimating the distribution of a quantity of interest from a sample with additive measurement error. Nearly all methods in the literature are based on Fourier transformation because it is mathematically a very neat solution. However, in practice these methods are unstable, and produce bad estimates when signal-noise ratio or sample size are low. In this paper, we develop a new deconvolution method based on maximum likelihood with a smoothness penalty. We show that our new method has much better performance than existing methods, particularly for small sample size or signal-noise ratio., Comment: 25 pages, 4 figures, Appendix - 30 pages 8 figures