Search

Sora's high-motion intensity and long consistent videos have significantly impacted the field of video generation, attracting unprecedented attention. However, existing publicly available datasets are inadequate for generating Sora-like videos, as they mainly contain short videos with low motion intensity and brief captions. To address these issues, we propose MiraData, a high-quality video dataset that surpasses previous ones in video duration, caption detail, motion strength, and visual quality. We curate MiraData from diverse, manually selected sources and meticulously process the data to obtain semantically consistent clips. GPT-4V is employed to annotate structured captions, providing detailed descriptions from four different perspectives along with a summarized dense caption. To better assess temporal consistency and motion intensity in video generation, we introduce MiraBench, which enhances existing benchmarks by adding 3D consistency and tracking-based motion strength metrics. MiraBench includes 150 evaluation prompts and 17 metrics covering temporal consistency, motion strength, 3D consistency, visual quality, text-video alignment, and distribution similarity. To demonstrate the utility and effectiveness of MiraData, we conduct experiments using our DiT-based video generation model, MiraDiT. The experimental results on MiraBench demonstrate the superiority of MiraData, especially in motion strength.

Filmmaking and animation production often require sophisticated techniques for coordinating camera transitions and object movements, typically involving labor-intensive real-world capturing. Despite advancements in generative AI for video creation, achieving precise control over motion for interactive video asset generation remains challenging. To this end, we propose Image Conductor, a method for precise control of camera transitions and object movements to generate video assets from a single image. An well-cultivated training strategy is proposed to separate distinct camera and object motion by camera LoRA weights and object LoRA weights. To further address cinematographic variations from ill-posed trajectories, we introduce a camera-free guidance technique during inference, enhancing object movements while eliminating camera transitions. Additionally, we develop a trajectory-oriented video motion data curation pipeline for training. Quantitative and qualitative experiments demonstrate our method's precision and fine-grained control in generating motion-controllable videos from images, advancing the practical application of interactive video synthesis. Project webpage available at https://liyaowei-stu.github.io/project/ImageConductor/, Comment: Project webpage available at https://liyaowei-stu.github.io/project/ImageConductor/

With the advancement of collaborative perception, the role of aerial-ground collaborative perception, a crucial component, is becoming increasingly important. The demand for collaborative perception across different perspectives to construct more comprehensive perceptual information is growing. However, challenges arise due to the disparities in the field of view between cross-domain agents and their varying sensitivity to information in images. Additionally, when we transform image features into Bird's Eye View (BEV) features for collaboration, we need accurate depth information. To address these issues, we propose a framework specifically designed for aerial-ground collaboration. First, to mitigate the lack of datasets for aerial-ground collaboration, we develop a virtual dataset named V2U-COO for our research. Second, we design a Cross-Domain Cross-Adaptation (CDCA) module to align the target information obtained from different domains, thereby achieving more accurate perception results. Finally, we introduce a Collaborative Depth Optimization (CDO) module to obtain more precise depth estimation results, leading to more accurate perception outcomes. We conduct extensive experiments on both our virtual dataset and a public dataset to validate the effectiveness of our framework. Our experiments on the V2U-COO dataset and the DAIR-V2X dataset demonstrate that our method improves detection accuracy by 6.1% and 2.7%, respectively.

Current instruction-based editing methods, such as InstructPix2Pix, often fail to produce satisfactory results in complex scenarios due to their dependence on the simple CLIP text encoder in diffusion models. To rectify this, this paper introduces SmartEdit, a novel approach to instruction-based image editing that leverages Multimodal Large Language Models (MLLMs) to enhance their understanding and reasoning capabilities. However, direct integration of these elements still faces challenges in situations requiring complex reasoning. To mitigate this, we propose a Bidirectional Interaction Module that enables comprehensive bidirectional information interactions between the input image and the MLLM output. During training, we initially incorporate perception data to boost the perception and understanding capabilities of diffusion models. Subsequently, we demonstrate that a small amount of complex instruction editing data can effectively stimulate SmartEdit's editing capabilities for more complex instructions. We further construct a new evaluation dataset, Reason-Edit, specifically tailored for complex instruction-based image editing. Both quantitative and qualitative results on this evaluation dataset indicate that our SmartEdit surpasses previous methods, paving the way for the practical application of complex instruction-based image editing., Comment: Project page: https://yuzhou914.github.io/SmartEdit/

Motions in a video primarily consist of camera motion, induced by camera movement, and object motion, resulting from object movement. Accurate control of both camera and object motion is essential for video generation. However, existing works either mainly focus on one type of motion or do not clearly distinguish between the two, limiting their control capabilities and diversity. Therefore, this paper presents MotionCtrl, a unified and flexible motion controller for video generation designed to effectively and independently control camera and object motion. The architecture and training strategy of MotionCtrl are carefully devised, taking into account the inherent properties of camera motion, object motion, and imperfect training data. Compared to previous methods, MotionCtrl offers three main advantages: 1) It effectively and independently controls camera motion and object motion, enabling more fine-grained motion control and facilitating flexible and diverse combinations of both types of motion. 2) Its motion conditions are determined by camera poses and trajectories, which are appearance-free and minimally impact the appearance or shape of objects in generated videos. 3) It is a relatively generalizable model that can adapt to a wide array of camera poses and trajectories once trained. Extensive qualitative and quantitative experiments have been conducted to demonstrate the superiority of MotionCtrl over existing methods. Project Page: https://wzhouxiff.github.io/projects/MotionCtrl/, Comment: SIGGRAPH 2024 Conference Proceedings

Incorporating a customized object into image generation presents an attractive feature in text-to-image generation. However, existing optimization-based and encoder-based methods are hindered by drawbacks such as time-consuming optimization, insufficient identity preservation, and a prevalent copy-pasting effect. To overcome these limitations, we introduce CustomNet, a novel object customization approach that explicitly incorporates 3D novel view synthesis capabilities into the object customization process. This integration facilitates the adjustment of spatial position relationships and viewpoints, yielding diverse outputs while effectively preserving object identity. Moreover, we introduce delicate designs to enable location control and flexible background control through textual descriptions or specific user-defined images, overcoming the limitations of existing 3D novel view synthesis methods. We further leverage a dataset construction pipeline that can better handle real-world objects and complex backgrounds. Equipped with these designs, our method facilitates zero-shot object customization without test-time optimization, offering simultaneous control over the viewpoints, location, and background. As a result, our CustomNet ensures enhanced identity preservation and generates diverse, harmonious outputs., Comment: Project webpage available at https://jiangyzy.github.io/CustomNet/

Fourier phase retrieval(PR) is a severely ill-posed inverse problem that arises in various applications. To guarantee a unique solution and relieve the dependence on the initialization, background information can be exploited as a structural priors. However, the requirement for the background information may be challenging when moving to the high-resolution imaging. At the same time, the previously proposed projected gradient descent(PGD) method also demands much background information. In this paper, we present an improved theoretical result about the demand for the background information, along with two Douglas Rachford(DR) based methods. Analytically, we demonstrate that the background required to ensure a unique solution can be decreased by nearly $1/2$ for the 2-D signals compared to the 1-D signals. By generalizing the results into $d$-dimension, we show that the length of the background information more than $(2^{\frac{d+1}{d}}-1)$ folds of the signal is sufficient to ensure the uniqueness. At the same time, we also analyze the stability and robustness of the model when measurements and background information are corrupted by the noise. Furthermore, two methods called Background Douglas-Rachford (BDR) and Convex Background Douglas-Rachford (CBDR) are proposed. BDR which is a kind of non-convex method is proven to have the local R-linear convergence rate under mild assumptions. Instead, CBDR method uses the techniques of convexification and can be proven to own a global convergence guarantee as long as the background information is sufficient. To support this, a new property called F-RIP is established. We test the performance of the proposed methods through simulations as well as real experimental measurements, and demonstrate that they achieve a higher recovery rate with less background information compared to the PGD method.

Fourier phase retrieval (FPR) is a challenging task widely used in various applications. It involves recovering an unknown signal from its Fourier phaseless measurements. FPR with few measurements is important for reducing time and hardware costs, but it suffers from serious ill-posedness. Recently, untrained neural networks have offered new approaches by introducing learned priors to alleviate the ill-posedness without requiring any external data. However, they may not be ideal for reconstructing fine details in images and can be computationally expensive. This paper proposes an untrained neural network (NN) embedded algorithm based on the alternating direction method of multipliers (ADMM) framework to solve FPR with few measurements. Specifically, we use a generative network to represent the image to be recovered, which confines the image to the space defined by the network structure. To improve the ability to represent high-frequency information, total variation (TV) regularization is imposed to facilitate the recovery of local structures in the image. Furthermore, to reduce the computational cost mainly caused by the parameter updates of the untrained NN, we develop an accelerated algorithm that adaptively trades off between explicit and implicit regularization. Experimental results indicate that the proposed algorithm outperforms existing untrained NN-based algorithms with fewer computational resources and even performs competitively against trained NN-based algorithms.

The randomized sparse Kaczmarz method, designed for seeking the sparse solutions of the linear systems $Ax=b$, selects the $i$-th projection hyperplane with likelihood proportional to $\|a_{i}\|_2^2$, where $a_{i}^T$ is $i$-th row of $A$. In this work, we propose a weighted randomized sparse Kaczmarz method, which selects the $i$-th projection hyperplane with probability proportional to $\lvert\langle a_{i},x_{k}\rangle-b_{i}\rvert^p$, where $0

3D GAN inversion aims to achieve high reconstruction fidelity and reasonable 3D geometry simultaneously from a single image input. However, existing 3D GAN inversion methods rely on time-consuming optimization for each individual case. In this work, we introduce a novel encoder-based inversion framework based on EG3D, one of the most widely-used 3D GAN models. We leverage the inherent properties of EG3D's latent space to design a discriminator and a background depth regularization. This enables us to train a geometry-aware encoder capable of converting the input image into corresponding latent code. Additionally, we explore the feature space of EG3D and develop an adaptive refinement stage that improves the representation ability of features in EG3D to enhance the recovery of fine-grained textural details. Finally, we propose an occlusion-aware fusion operation to prevent distortion in unobserved regions. Our method achieves impressive results comparable to optimization-based methods while operating up to 500 times faster. Our framework is well-suited for applications such as semantic editing., Comment: Project page: https://eg3d-goae.github.io/

Fourier phase retrieval (FPR) is an inverse problem that recovers the signal from its Fourier magnitude measurement, it's ill-posed especially when the sampling rates are low. In this paper, an untrained generative prior is introduced to attack the ill-posedness. Based on the alternating direction method of multipliers (ADMM), an algorithm utilizing the untrained generative network called Net-ADM is proposed to solve the FPR problem. Firstly, the objective function is smoothed and the dimension of the variable is raised to facilitate calculation. Then an untrained generative network is embedded in the iterative process of ADMM to project an estimated signal into the generative space, and the projected signal is applied to next iteration of ADMM. We theoretically analyzed the two projections included in the algorithm, one makes the objective function descent, and the other gets the estimation closer to the optimal solution. Numerical experiments show that the reconstruction performance and robustness of the proposed algorithm are superior to prior works, especially when the sampling rates are low.

Free-form text prompts allow users to describe their intentions during image manipulation conveniently. Based on the visual latent space of StyleGAN[21] and text embedding space of CLIP[34], studies focus on how to map these two latent spaces for text-driven attribute manipulations. Currently, the latent mapping between these two spaces is empirically designed and confines that each manipulation model can only handle one fixed text prompt. In this paper, we propose a method named Free-Form CLIP (FFCLIP), aiming to establish an automatic latent mapping so that one manipulation model handles free-form text prompts. Our FFCLIP has a cross-modality semantic modulation module containing semantic alignment and injection. The semantic alignment performs the automatic latent mapping via linear transformations with a cross attention mechanism. After alignment, we inject semantics from text prompt embeddings to the StyleGAN latent space. For one type of image (e.g., `human portrait'), one FFCLIP model can be learned to handle free-form text prompts. Meanwhile, we observe that although each training text prompt only contains a single semantic meaning, FFCLIP can leverage text prompts with multiple semantic meanings for image manipulation. In the experiments, we evaluate FFCLIP on three types of images (i.e., `human portraits', `cars', and `churches'). Both visual and numerical results show that FFCLIP effectively produces semantically accurate and visually realistic images. Project page: https://github.com/KumapowerLIU/FFCLIP., Comment: Accepted by NeurIPS 2022

Light-matter interaction is exploited in spectroscopic techniques to access information about molecular, atomic or nuclear constituents of the sample of interest. While scattered light carries both amplitude and phase information of the electromagnetic field, most of the time the latter is lost in intensity measurements. However, often the phase information is paramount to reconstruct the desired information of the target, as it is well known from coherent x-ray imaging. Here we introduce a new phase retrieval algorithm which allows us to reconstruct the field phase information from two-dimensional time- and energy-resolved spectra. We apply this method to the particular case of x-ray scattering off M\"ossbauer nuclei at a synchrotron radiation source. Knowledge of the phase allows also for an excellent reconstruction of the energy spectra from experimental data, which could not be achieved with this resolution otherwise. Our approach provides an efficient novel data analysis tool which will benefit x-ray quantum optics and M\"ossbauer spectroscopy with synchrotron radiation alike.

The sketch-and-project, as a general archetypal algorithm for solving linear systems, unifies a variety of randomized iterative methods such as the randomized Kaczmarz and randomized coordinate descent. However, since it aims to find a least-norm solution from a linear system, the randomized sparse Kaczmarz can not be included. This motivates us to propose a more general framework, called sketched Bregman projection (SBP) method, in which we are able to find solutions with certain structures from linear systems. To generalize the concept of adaptive sampling to the SBP method, we show how the progress, measured by Bregman distance, of single step depends directly on a sketched loss function. Theoretically, we provide detailed global convergence results for the SBP method with different adaptive sampling rules. At last, for the (sparse) Kaczmarz methods, a group of numerical simulations are tested, with which we verify that the methods utilizing sampling Kaczmarz-Motzkin rule demands the fewest computational costs to achieve a given error bound comparing to the corresponding methods with other sampling rules.

The randomized sparse Kaczmarz method was recently proposed to recover sparse solutions of linear systems. In this work, we introduce a greedy variant of the randomized sparse Kaczmarz method by employing the sampling Kaczmarz-Motzkin method, and prove its linear convergence in expectation with respect to the Bregman distance in the noiseless and noisy cases. This greedy variant can be viewed as a unification of the sampling Kaczmarz-Motzkin method and the randomized sparse Kaczmarz method, and hence inherits the merits of these two methods. Numerically, we report a couple of experimental results to demonstrate its superiority

In this paper, a Parallel Direct Eigensolver for Sequences of Hermitian Eigenvalue Problems with no tridiagonalization is proposed, denoted by \texttt{PDESHEP}, and it combines direct methods with iterative methods. \texttt{PDESHEP} first reduces a Hermitian matrix to its banded form, then applies a spectrum slicing algorithm to the banded matrix, and finally computes the eigenvectors of the original matrix via backtransform. Therefore, compared with conventional direct eigensolvers, \texttt{PDESHEP} avoids tridiagonalization, which consists of many memory-bounded operations. In this work, the iterative method in \texttt{PDESHEP} is based on the contour integral method implemented in FEAST. The combination of direct methods with iterative methods for banded matrices requires some efficient data redistribution algorithms both from 2D to 1D and from 1D to 2D data structures. Hence, some two-step data redistribution algorithms are proposed, which can be $10\times$ faster than ScaLAPACK routine \texttt{PXGEMR2D}. For the symmetric self-consistent field (SCF) eigenvalue problems, \texttt{PDESHEP} can be on average $1.25\times$ faster than the state-of-the-art direct solver in ELPA when using $4096$ processes. Numerical results are obtained for dense Hermitian matrices from real applications and large real sparse matrices from the SuiteSparse collection., Comment: 19 pages and 9 figures

Change detection is a basic task of remote sensing image processing. The research objective is to identity the change information of interest and filter out the irrelevant change information as interference factors. Recently, the rise of deep learning has provided new tools for change detection, which have yielded impressive results. However, the available methods focus mainly on the difference information between multitemporal remote sensing images and lack robustness to pseudo-change information. To overcome the lack of resistance of current methods to pseudo-changes, in this paper, we propose a new method, namely, dual attentive fully convolutional Siamese networks (DASNet) for change detection in high-resolution images. Through the dual-attention mechanism, long-range dependencies are captured to obtain more discriminant feature representations to enhance the recognition performance of the model. Moreover, the imbalanced sample is a serious problem in change detection, i.e. unchanged samples are much more than changed samples, which is one of the main reasons resulting in pseudo-changes. We put forward the weighted double margin contrastive loss to address this problem by punishing the attention to unchanged feature pairs and increase attention to changed feature pairs. The experimental results of our method on the change detection dataset (CDD) and the building change detection dataset (BCDD) demonstrate that compared with other baseline methods, the proposed method realizes maximum improvements of 2.1\% and 3.6\%, respectively, in the F1 score. Our Pytorch implementation is available at https://github.com/lehaifeng/DASNet., Comment: 12 pages, 13 figures, 5 tables

Recovering the image of an object from its phaseless speckle pattern is difficult. Let alone the transmission matrix is unknown in multiple scattering media imaging. Double phase retrieval is a recently proposed efficient method which recovers the unknown object from its phaseless measurements by two steps with phase retrieval. In this paper, we combine the two steps in double phase retrieval and construct an end-to-end neural network called TCNN(Transforming Convolutional Neural Network) which directly learns the relationship between the phaseless measurements and the object. TCNN contains a special layer called transform layer which aims to be a bridge between different transform domains. Tested by the empirical data provided in\cite{Metzler2017Coherent}, images can be recovered by TCNN with comparable quality compared with state-of-the-art methods. Not only the transmission matrix needn't to be calculated but also the time to recover the object can be hugely reduced once the parameters of TCNN are stable.

Phase retrieval problem has been studied in various applications. It is an inverse problem without the standard uniqueness guarantee. To make complete theoretical analyses and devise efficient algorithms to recover the signal is sophisticated. In this paper, we come up with a model called \textit{phase retrieval with background information} which recovers the signal with the known background information from the intensity of their combinational Fourier transform spectrum. We prove that the uniqueness of phase retrieval can be guaranteed even considering those trivial solutions when the background information is sufficient. Under this condition, we construct a loss function and utilize the projected gradient descent method to search for the ground truth. We prove that the stationary point is the global optimum with probability 1. Numerical simulations demonstrate the projected gradient descent method performs well both for 1-D and 2-D signals. Furthermore, this method is quite robust to the Gaussian noise and the bias of the background information.

Phase retrieval(PR) problem is a kind of ill-condition inverse problem which can be found in various of applications. Utilizing the sparse priority, an algorithm called SWF(Sparse Wirtinger Flow) is proposed in this paper to deal with sparse PR problem based on the Wirtinger flow method. SWF firstly recovers the support of the signal and then updates the evaluation by hard thresholding method with an elaborate initialization. Theoretical analyses show that SWF has a geometric convergence for any $k$ sparse $n$ length signal with the sampling complexity $\mathcal{O}(k^2\mathrm{log}n)$. To get $\varepsilon$ accuracy, the computational complexity of SWF is $\mathcal{O}(k^3n\mathrm{log}n\mathrm{log}\frac{1}{\varepsilon})$. Numerical tests also demonstrate that SWF performs better than state-of-the-art methods especially when we have no priori knowledge about sparsity $k$. Moreover, SWF is also robust to the noise

Phase retrieval(PR) problem is a kind of ill-condition inverse problem which is arising in various of applications. Based on the Wirtinger flow(WF) method, a reweighted Wirtinger flow(RWF) method is proposed to deal with PR problem. RWF finds the global optimum by solving a series of sub-PR problems with changing weights. Theoretical analyses illustrate that the RWF has a geometric convergence from a deliberate initialization when the weights are bounded by 1 and $\frac{10}{9}$. Numerical testing shows RWF has a lower sampling complexity compared with WF. As an essentially adaptive truncated Wirtinger flow(TWF) method, RWF performs better than TWF especially when the ratio between sampling number $m$ and length of signal $n$ is small., Comment: 21 pages, 11 figures