PyTorchKR Papers

최신 논문

총 3,696편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2209.14792

Make-A-Video: Text-to-Video Generation without Text-Video Data

We propose Make-A-Video -- an approach for directly translating the tremendous recent progress in Text-to-Image (T2I) generation to Text-to-Video (T2V). Our intuition is simple: learn what the world looks like and how it is described from paired text-image data, and learn how the world moves from unsupervised video footage. Make-A-Video has three advantages: (1) it accelerates training of the T2V model (it does not need to learn visual and multimodal representations from scratch), (2) it does not require paired text-video data, and (3) the generated videos inherit the vastness (diversity in aesthetic, fantastical depictions, etc.) of today's image generation models. We design a simple yet effective way to build on T2I models with novel and effective spatial-temporal modules. First, we decompose the full temporal U-Net and attention tensors and approximate them in space and time. Second, we design a spatial temporal pipeline to generate high resolution and frame rate videos with a video decoder, interpolation model and two super resolution models that can enable various applications besides T2V. In all aspects, spatial and temporal resolution, faithfulness to text, and quality, Make-A-Video sets the new state-of-the-art in text-to-video generation, as determined by both qualitative and quantitative measures.

arXiv:2209.14145공식 코드 · Papers with Code

Multi-scale Attention Network for Single Image Super-Resolution

ConvNets can compete with transformers in high-level tasks by exploiting larger receptive fields. To unleash the potential of ConvNet in super-resolution, we propose a multi-scale attention network (MAN), by coupling classical multi-scale mechanism with emerging large kernel attention. In particular, we proposed multi-scale large kernel attention (MLKA) and gated spatial attention unit (GSAU). Through our MLKA, we modify large kernel attention with multi-scale and gate schemes to obtain the abundant attention map at various granularity levels, thereby aggregating global and local information and avoiding potential blocking artifacts. In GSAU, we integrate gate mechanism and spatial attention to remove the unnecessary linear layer and aggregate informative spatial context. To confirm the effectiveness of our designs, we evaluate MAN with multiple complexities by simply stacking different numbers of MLKA and GSAU. Experimental results illustrate that our MAN can perform on par with SwinIR and achieve varied trade-offs between state-of-the-art performance and computations.

arXiv:2209.13131

Hitchhiker's Guide to Super-Resolution: Introduction and Recent Advances

With the advent of Deep Learning (DL), Super-Resolution (SR) has also become a thriving research area. However, despite promising results, the field still faces challenges that require further research e.g., allowing flexible upsampling, more effective loss functions, and better evaluation metrics. We review the domain of SR in light of recent advances, and examine state-of-the-art models such as diffusion (DDPM) and transformer-based SR models. We present a critical discussion on contemporary strategies used in SR, and identify promising yet unexplored research directions. We complement previous surveys by incorporating the latest developments in the field such as uncertainty-driven losses, wavelet networks, neural architecture search, novel normalization methods, and the latests evaluation techniques. We also include several visualizations for the models and methods throughout each chapter in order to facilitate a global understanding of the trends in the field. This review is ultimately aimed at helping researchers to push the boundaries of DL applied to SR.

arXiv:2209.13055

Effective Invertible Arbitrary Image Rescaling

Great successes have been achieved using deep learning techniques for image super-resolution (SR) with fixed scales. To increase its real world applicability, numerous models have also been proposed to restore SR images with arbitrary scale factors, including asymmetric ones where images are resized to different scales along horizontal and vertical directions. Though most models are only optimized for the unidirectional upscaling task while assuming a predefined downscaling kernel for low-resolution (LR) inputs, recent models based on Invertible Neural Networks (INN) are able to increase upscaling accuracy significantly by optimizing the downscaling and upscaling cycle jointly. However, limited by the INN architecture, it is constrained to fixed integer scale factors and requires one model for each scale. Without increasing model complexity, a simple and effective invertible arbitrary rescaling network (IARN) is proposed to achieve arbitrary image rescaling by training only one model in this work. Using innovative components like position-aware scale encoding and preemptive channel splitting, the network is optimized to convert the non-invertible rescaling cycle to an effectively invertible process. It is shown to achieve a state-of-the-art (SOTA) performance in bidirectional arbitrary rescaling without compromising perceptual quality in LR outputs. It is also demonstrated to perform well on tests with asymmetric scales using the same network architecture.

arXiv:2209.13362

DELTAR: Depth Estimation from a Light-weight ToF Sensor and RGB Image

Light-weight time-of-flight (ToF) depth sensors are small, cheap, low-energy and have been massively deployed on mobile devices for the purposes like autofocus, obstacle detection, etc. However, due to their specific measurements (depth distribution in a region instead of the depth value at a certain pixel) and extremely low resolution, they are insufficient for applications requiring high-fidelity depth such as 3D reconstruction. In this paper, we propose DELTAR, a novel method to empower light-weight ToF sensors with the capability of measuring high resolution and accurate depth by cooperating with a color image. As the core of DELTAR, a feature extractor customized for depth distribution and an attention-based neural architecture is proposed to fuse the information from the color and ToF domain efficiently. To evaluate our system in real-world scenarios, we design a data collection device and propose a new approach to calibrate the RGB camera and ToF sensor. Experiments show that our method produces more accurate depth than existing frameworks designed for depth completion and depth super-resolution and achieves on par performance with a commodity-level RGB-D sensor. Code and data are available at https://zju3dv.github.io/deltar/.

arXiv:2209.13435공식 코드 · Papers with Code

Scaling Laws For Deep Learning Based Image Reconstruction

Deep neural networks trained end-to-end to map a measurement of a (noisy) image to a clean image perform excellent for a variety of linear inverse problems. Current methods are only trained on a few hundreds or thousands of images as opposed to the millions of examples deep networks are trained on in other domains. In this work, we study whether major performance gains are expected from scaling up the training set size. We consider image denoising, accelerated magnetic resonance imaging, and super-resolution and empirically determine the reconstruction quality as a function of training set size, while simultaneously scaling the network size. For all three tasks we find that an initially steep power-law scaling slows significantly already at moderate training set sizes. Interpolating those scaling laws suggests that even training on millions of images would not significantly improve performance. To understand the expected behavior, we analytically characterize the performance of a linear estimator learned with early stopped gradient descent. The result formalizes the intuition that once the error induced by learning the signal model is small relative to the error floor, more training examples do not improve performance.

arXiv:2209.13351공식 코드 · Papers with Code

SuperYOLO: Super Resolution Assisted Object Detection in Multimodal Remote Sensing Imagery

Accurately and timely detecting multiscale small objects that contain tens of pixels from remote sensing images (RSI) remains challenging. Most of the existing solutions primarily design complex deep neural networks to learn strong feature representations for objects separated from the background, which often results in a heavy computation burden. In this article, we propose an accurate yet fast object detection method for RSI, named SuperYOLO, which fuses multimodal data and performs high-resolution (HR) object detection on multiscale objects by utilizing the assisted super resolution (SR) learning and considering both the detection accuracy and computation cost. First, we utilize a symmetric compact multimodal fusion (MF) to extract supplementary information from various data for improving small object detection in RSI. Furthermore, we design a simple and flexible SR branch to learn HR feature representations that can discriminate small objects from vast backgrounds with low-resolution (LR) input, thus further improving the detection accuracy. Moreover, to avoid introducing additional computation, the SR branch is discarded in the inference stage, and the computation of the network model is reduced due to the LR input. Experimental results show that, on the widely used VEDAI RS dataset, SuperYOLO achieves an accuracy of 75.09% (in terms of mAP50 ), which is more than 10% higher than the SOTA large models, such as YOLOv5l, YOLOv5x, and RS designed YOLOrs. Meanwhile, the parameter size and GFLOPs of SuperYOLO are about 18 times and 3.8 times less than YOLOv5x. Our proposed model shows a favorable accuracy and speed tradeoff compared to the state-of-the-art models. The code will be open-sourced at https://github.com/icey-zhang/SuperYOLO.

arXiv:2209.12875

Realistic Hair Synthesis with Generative Adversarial Networks

Recent successes in generative modeling have accelerated studies on this subject and attracted the attention of researchers. One of the most important methods used to achieve this success is Generative Adversarial Networks (GANs). It has many application areas such as; virtual reality (VR), augmented reality (AR), super resolution, image enhancement. Despite the recent advances in hair synthesis and style transfer using deep learning and generative modelling, due to the complex nature of hair still contains unsolved challenges. The methods proposed in the literature to solve this problem generally focus on making high-quality hair edits on images. In this thesis, a generative adversarial network method is proposed to solve the hair synthesis problem. While developing this method, it is aimed to achieve real-time hair synthesis while achieving visual outputs that compete with the best methods in the literature. The proposed method was trained with the FFHQ dataset and then its results in hair style transfer and hair reconstruction tasks were evaluated. The results obtained in these tasks and the operating time of the method were compared with MichiGAN, one of the best methods in the literature. The comparison was made at a resolution of 128x128. As a result of the comparison, it has been shown that the proposed method achieves competitive results with MichiGAN in terms of realistic hair synthesis, and performs better in terms of operating time.

arXiv:2209.12433

Deep generative model super-resolves spatially correlated multiregional climate data

Super-resolving the coarse outputs of global climate simulations, termed downscaling, is crucial in making political and social decisions on systems requiring long-term climate change projections. Existing fast super-resolution techniques, however, have yet to preserve the spatially correlated nature of climatological data, which is particularly important when we address systems with spatial expanse, such as the development of transportation infrastructure. Herein, we show an adversarial network-based machine learning enables us to correctly reconstruct the inter-regional spatial correlations in downscaling with high magnification of up to fifty while maintaining pixel-wise statistical consistency. Direct comparison with the measured meteorological data of temperature and precipitation distributions reveals that integrating climatologically important physical information improves the downscaling performance, which prompts us to call this approach $\pi$SRGAN (Physics Informed Super-Resolution Generative Adversarial Network). The proposed method has a potential application to the inter-regionally consistent assessment of the climate change impact. Additionally, we present the outcomes of another variant of the deep generative model-based downscaling approach in which the low-resolution precipitation field is substituted with the pressure field, referred to as $\psi$SRGAN (Precipitation Source Inaccessible SRGAN). Remarkably, this method demonstrates unexpectedly good downscaling performance for the precipitation field.

arXiv:2209.12406공식 코드 · Papers with Code

A heterogeneous group CNN for image super-resolution

Convolutional neural networks (CNNs) have obtained remarkable performance via deep architectures. However, these CNNs often achieve poor robustness for image super-resolution (SR) under complex scenes. In this paper, we present a heterogeneous group SR CNN (HGSRCNN) via leveraging structure information of different types to obtain a high-quality image. Specifically, each heterogeneous group block (HGB) of HGSRCNN uses a heterogeneous architecture containing a symmetric group convolutional block and a complementary convolutional block in a parallel way to enhance internal and external relations of different channels for facilitating richer low-frequency structure information of different types. To prevent appearance of obtained redundant features, a refinement block with signal enhancements in a serial way is designed to filter useless information. To prevent loss of original information, a multi-level enhancement mechanism guides a CNN to achieve a symmetric architecture for promoting expressive ability of HGSRCNN. Besides, a parallel up-sampling mechanism is developed to train a blind SR model. Extensive experiments illustrate that the proposed HGSRCNN has obtained excellent SR performance in terms of both quantitative and qualitative analysis. Codes can be accessed at https://github.com/hellloxiaotian/HGSRCNN.

arXiv:2209.12475공식 코드 · Papers with Code

Real-RawVSR: Real-World Raw Video Super-Resolution with a Benchmark Dataset

In recent years, real image super-resolution (SR) has achieved promising results due to the development of SR datasets and corresponding real SR methods. In contrast, the field of real video SR is lagging behind, especially for real raw videos. Considering the superiority of raw image SR over sRGB image SR, we construct a real-world raw video SR (Real-RawVSR) dataset and propose a corresponding SR method. We utilize two DSLR cameras and a beam-splitter to simultaneously capture low-resolution (LR) and high-resolution (HR) raw videos with 2x, 3x, and 4x magnifications. There are 450 video pairs in our dataset, with scenes varying from indoor to outdoor, and motions including camera and object movements. To our knowledge, this is the first real-world raw VSR dataset. Since the raw video is characterized by the Bayer pattern, we propose a two-branch network, which deals with both the packed RGGB sequence and the original Bayer pattern sequence, and the two branches are complementary to each other. After going through the proposed co-alignment, interaction, fusion, and reconstruction modules, we generate the corresponding HR sRGB sequence. Experimental results demonstrate that the proposed method outperforms benchmark real and synthetic video SR methods with either raw or sRGB inputs. Our code and dataset are available at https://github.com/zmzhang1998/Real-RawVSR.

arXiv:2209.12064공식 코드 · Papers with Code

Face Super-Resolution Using Stochastic Differential Equations

Diffusion models have proven effective for various applications such as images, audio and graph generation. Other important applications are image super-resolution and the solution of inverse problems. More recently, some works have used stochastic differential equations (SDEs) to generalize diffusion models to continuous time. In this work, we introduce SDEs to generate super-resolution face images. To the best of our knowledge, this is the first time SDEs have been used for such an application. The proposed method provides an improved peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), and consistency than the existing super-resolution methods based on diffusion models. In particular, we also assess the potential application of this method for the face recognition task. A generic facial feature extractor is used to compare the super-resolution images with the ground truth and superior results were obtained compared with other methods. Our code is publicly available at https://github.com/marcelowds/sr-sde

arXiv:2209.11345공식 코드 · Papers with Code

Swin2SR: SwinV2 Transformer for Compressed Image Super-Resolution and Restoration

Compression plays an important role on the efficient transmission and storage of images and videos through band-limited systems such as streaming services, virtual reality or videogames. However, compression unavoidably leads to artifacts and the loss of the original information, which may severely degrade the visual quality. For these reasons, quality enhancement of compressed images has become a popular research topic. While most state-of-the-art image restoration methods are based on convolutional neural networks, other transformers-based methods such as SwinIR, show impressive performance on these tasks. In this paper, we explore the novel Swin Transformer V2, to improve SwinIR for image super-resolution, and in particular, the compressed input scenario. Using this method we can tackle the major issues in training transformer vision models, such as training instability, resolution gaps between pre-training and fine-tuning, and hunger on data. We conduct experiments on three representative tasks: JPEG compression artifacts removal, image super-resolution (classical and lightweight), and compressed image super-resolution. Experimental results demonstrate that our method, Swin2SR, can improve the training convergence and performance of SwinIR, and is a top-5 solution at the "AIM 2022 Challenge on Super-Resolution of Compressed Image and Video".

논문 ID: document/9511400

Crack Segmentation for Low-Resolution Images using Joint Learning with Super-Resolution

This paper proposes a method for crack segmentation on low-resolution images. Detailed cracks on their high-resolution images are estimated by super resolution from the low-resolution images. Our proposed method optimizes super-resolution images for the crack segmentation. For this method, we propose the Boundary Combo loss to express the local details of the crack. Experimental results demonstrate that our method outperforms the combinations of other previous approaches.

논문 ID: 4085.34752

Boosting Lightweight Single Image Super-resolution via Joint-distillation

The rising of deep learning has facilitated the development of single image super-resolution (SISR). However, the growing burdensome model complexity and memory occupation severely hinder its practical deployments on resource-limited devices. In this paper, we propose a novel joint-distillation (JDSR) framework to boost the representation of various off-the-shelf lightweight SR models. The framework includes two stages: the superior LR generation and the joint-distillation learning. The superior LR is obtained from the HR image itself. With less than $300$K parameters, the peer network using superior LR as input can achieve comparable SR performance with large models, e.g., RCAN, with 15M parameters, which enables it as the input of peer network to save the training expense. The joint-distillation learning consists of internal self-distillation and external mutual learning. The internal self-distillation aims to achieve model self-boosting by transferring the knowledge from the deeper SR output to the shallower one. Specifically, each intermediate SR output is supervised by the HR image and the soft label from subsequent deeper outputs. To shrink the capacity gap between shallow and deep layers, a soft label generator is designed in a progressive backward fusion way with meta-learning for adaptive weight fine-tuning. The external mutual learning focuses on obtaining interaction information from a peer network in the process. Moreover, a curriculum learning strategy and a performance gap threshold are introduced for balancing the convergence rate of the original SR model and its peer network. Comprehensive experiments on benchmark datasets demonstrate that our proposal improves the performance of recent lightweight SR models by a large margin, with the same model architecture and inference expense.

arXiv:2209.11037

3D Super-Resolution Imaging Method for Distributed Millimeter-wave Automotive Radar System

Millimeter-wave (mmW) radar is widely applied to advanced autopilot assistance systems. However, its small antenna aperture causes a low imaging resolution. In this paper, a new distributed mmW radar system is designed to solve this problem. It forms a large sparse virtual planar array to enlarge the aperture, using multiple-input and multiple-output (MIMO) processing. However, in this system, traditional imaging methods cannot apply to the sparse array. Therefore, we also propose a 3D super-resolution imaging method specifically for this system in this paper. The proposed method consists of three steps: (1) using range FFT to get range imaging, (2) using 2D adaptive diagonal loading iterative adaptive approach (ADL-IAA) to acquire 2D super-resolution imaging, which can satisfy this sparsity under single-measurement, (3) using constant false alarm (CFAR) processing to gain final 3D super-resolution imaging. The simulation results show the proposed method can significantly improve imaging resolution under the sparse array and single-measurement.

arXiv:2209.11112공식 코드 · Papers with Code

CMGAN: Conformer-Based Metric-GAN for Monaural Speech Enhancement

In this work, we further develop the conformer-based metric generative adversarial network (CMGAN) model for speech enhancement (SE) in the time-frequency (TF) domain. This paper builds on our previous work but takes a more in-depth look by conducting extensive ablation studies on model inputs and architectural design choices. We rigorously tested the generalization ability of the model to unseen noise types and distortions. We have fortified our claims through DNS-MOS measurements and listening tests. Rather than focusing exclusively on the speech denoising task, we extend this work to address the dereverberation and super-resolution tasks. This necessitated exploring various architectural changes, specifically metric discriminator scores and masking techniques. It is essential to highlight that this is among the earliest works that attempted complex TF-domain super-resolution. Our findings show that CMGAN outperforms existing state-of-the-art methods in the three major speech enhancement tasks: denoising, dereverberation, and super-resolution. For example, in the denoising task using the Voice Bank+DEMAND dataset, CMGAN notably exceeded the performance of prior models, attaining a PESQ score of 3.41 and an SSNR of 11.10 dB. Audio samples and CMGAN implementations are available online.

arXiv:2209.10057

Analytic Optimization-Based Microbubble Tracking in Ultrasound Super-Resolution Microscopy

Ultrasound localization microscopy (ULM) refers to a promising medical imaging modality that systematically leverages the advantages of contrast-enhanced ultrasound (CEUS) to surpass the diffraction barrier and delineate the microvascular map. Localization and tracking of microbubbles (MBs), two significant steps of ULM, facilitate generating the vascular map and the velocity distribution, respectively. Herein, we propose a novel MB tracking technique considering temporal pairing as a bubble-set registration problem. Iterative registration is performed between the bubble sets in two consecutive time instants by analytically optimizing a cost function that takes position and point-spread function (PSF) similarities as well as physically plausible levels of bubbles' movement into account. Furthermore, we infer MBs' parity in a fuzzy manner instead of binary. The proposed technique performs well in validation experiments with two synthetic and two in vivo datasets provided by the Ultrasound Localisation and TRacking Algorithms for Super Resolution (ULTRA-SR) Challenge.

arXiv:2209.10305공식 코드 · Papers with Code

KXNet: A Model-Driven Deep Neural Network for Blind Super-Resolution

Although current deep learning-based methods have gained promising performance in the blind single image super-resolution (SISR) task, most of them mainly focus on heuristically constructing diverse network architectures and put less emphasis on the explicit embedding of the physical generation mechanism between blur kernels and high-resolution (HR) images. To alleviate this issue, we propose a model-driven deep neural network, called KXNet, for blind SISR. Specifically, to solve the classical SISR model, we propose a simple-yet-effective iterative algorithm. Then by unfolding the involved iterative steps into the corresponding network module, we naturally construct the KXNet. The main specificity of the proposed KXNet is that the entire learning process is fully and explicitly integrated with the inherent physical mechanism underlying this SISR task. Thus, the learned blur kernel has clear physical patterns and the mutually iterative process between blur kernel and HR image can soundly guide the KXNet to be evolved in the right direction. Extensive experiments on synthetic and real data finely demonstrate the superior accuracy and generality of our method beyond the current representative state-of-the-art blind SISR methods. Code is available at: https://github.com/jiahong-fu/KXNet.

arXiv:2209.10192

Multi-Field De-interlacing using Deformable Convolution Residual Blocks and Self-Attention

Although deep learning has made significant impact on image/video restoration and super-resolution, learned deinterlacing has so far received less attention in academia or industry. This is despite deinterlacing is well-suited for supervised learning from synthetic data since the degradation model is known and fixed. In this paper, we propose a novel multi-field full frame-rate deinterlacing network, which adapts the state-of-the-art superresolution approaches to the deinterlacing task. Our model aligns features from adjacent fields to a reference field (to be deinterlaced) using both deformable convolution residual blocks and self attention. Our extensive experimental results demonstrate that the proposed method provides state-of-the-art deinterlacing results in terms of both numerical and perceptual performance. At the time of writing, our model ranks first in the Full FrameRate LeaderBoard at https://videoprocessing.ai/benchmarks/deinterlacer.html