视频对象细分（VOS）是视频理解的基础。基于变压器的方法在半监督VOS上显示出显着的性能改善。但是，现有的工作面临着挑战在彼此近距离接近视觉上类似对象的挑战。在本文中，我们提出了一种新型的双边注意力变压器，以进行半监督VO的运动出现空间（蝙蝠侠）。它通过新型的光流校准模块在视频中捕获对象运动，该模块将分割面膜与光流估计融合在一起，以改善对象内光流平滑度并减少物体边界处的噪声。然后在我们的新型双边注意力中采用了这种校准的光流，该流动流在相邻双边空间中的查询和参考帧之间的对应关系考虑，考虑到运动和外观。广泛的实验通过在所有四个流行的VOS基准上胜过所有现有最新的实验：YouTube-VOS 2019（85.0％），YouTube-VOS 2018（85.3％），Davis 2017VAL/TESTDEV（86.2.2 ％/82.2％）和戴维斯（Davis）2016（92.5％）。

在线操作检测是一旦在流视频中进行的操作，就可以预测该动作。一个主要的挑战是，该模型无法访问未来，并且必须仅依靠历史，即到目前为止观察到的框架来做出预测。因此，重要的是要强调历史的一部分，这些部分对当前框架的预测更有意义。我们提出了带有背景抑制的封闭历史单元的Gatehub，其中包括一种新颖的位置引导的封闭式跨注意机制，以增强或抑制历史的一部分，因为它们在当前框架预测方面的信息程度。 GateHub进一步建议未来的历史记录（FAH），通过使用后来观察到的框架，使历史特征更具信息性。在一个统一的框架中，GateHub集成了变压器的远程时间建模的能力以及经常性模型选择性编码相关信息的能力。 GateHub还引入了一个背景抑制目标，以进一步减轻与动作框架非常相似的误报背景框架。对三个基准数据集（Thumos，TVSeries和HDD）进行了广泛的验证，这表明GateHub显着胜过所有现有方法，并且比现有最佳工作更有效。此外，与所有需要RGB和光流信息进行预测的现有方法相比，GateHub的无流版本能够以2.8倍的帧速率获得更高或密切的精度。

Mixup is a popular data augmentation technique for training deep neural networks where additional samples are generated by linearly interpolating pairs of inputs and their labels. This technique is known to improve the generalization performance in many learning paradigms and applications. In this work, we first analyze Mixup and show that it implicitly regularizes infinitely many directional derivatives of all orders. We then propose a new method to improve Mixup based on the novel insight. To demonstrate the effectiveness of the proposed method, we conduct experiments across various domains such as images, tabular data, speech, and graphs. Our results show that the proposed method improves Mixup across various datasets using a variety of architectures, for instance, exhibiting an improvement over Mixup by 0.8% in ImageNet top-1 accuracy.

This paper presents a comprehensive survey of low-light image and video enhancement. We begin with the challenging mixed over-/under-exposed images, which are under-performed by existing methods. To this end, we propose two variants of the SICE dataset named SICE_Grad and SICE_Mix. Next, we introduce Night Wenzhou, a large-scale, high-resolution video dataset, to address the issue of the lack of a low-light video dataset that discount the use of low-light image enhancement (LLIE) to videos. The Night Wenzhou dataset is challenging since it consists of fast-moving aerial scenes and streetscapes with varying illuminations and degradation. We conduct extensive key technique analysis and experimental comparisons for representative LLIE approaches using these newly proposed datasets and the current benchmark datasets. Finally, we address unresolved issues and propose future research topics for the LLIE community.

We investigate data-driven texture modeling via analysis and synthesis with generative adversarial networks. For network training and testing, we have compiled a diverse set of spatially homogeneous textures, ranging from stochastic to regular. We adopt StyleGAN3 for synthesis and demonstrate that it produces diverse textures beyond those represented in the training data. For texture analysis, we propose GAN inversion using a novel latent domain reconstruction consistency criterion for synthesized textures, and iterative refinement with Gramian loss for real textures. We propose perceptual procedures for evaluating network capabilities, exploring the global and local behavior of latent space trajectories, and comparing with existing texture analysis-synthesis techniques.

Recent advances in deep learning research, such as transformers, have bolstered the ability for automated agents to generate creative texts similar to those that a human would write. By default, transformer decoders can only generate new text with respect to previously generated text. The output distribution of candidate tokens at any position is conditioned on previously selected tokens using a self-attention mechanism to emulate the property of autoregression. This is inherently limiting for tasks such as controllable story generation where it may be necessary to condition on future plot events when writing a story. In this work, we propose Future Sight, a method for finetuning a pretrained generative transformer on the task of future conditioning. Transformer decoders are typically pretrained on the task of completing a context, one token at a time, by means of self-attention. Future Sight additionally enables a decoder to attend to an encoded future plot event. This motivates the decoder to expand on the context in a way that logically concludes with the provided future. During inference, the future plot event can be written by a human author to steer the narrative being generated in a certain direction. We evaluate the efficacy of our approach on a story generation task with human evaluators.

Transformer-based models have gained large popularity and demonstrated promising results in long-term time-series forecasting in recent years. In addition to learning attention in time domain, recent works also explore learning attention in frequency domains (e.g., Fourier domain, wavelet domain), given that seasonal patterns can be better captured in these domains. In this work, we seek to understand the relationships between attention models in different time and frequency domains. Theoretically, we show that attention models in different domains are equivalent under linear conditions (i.e., linear kernel to attention scores). Empirically, we analyze how attention models of different domains show different behaviors through various synthetic experiments with seasonality, trend and noise, with emphasis on the role of softmax operation therein. Both these theoretical and empirical analyses motivate us to propose a new method: TDformer (Trend Decomposition Transformer), that first applies seasonal-trend decomposition, and then additively combines an MLP which predicts the trend component with Fourier attention which predicts the seasonal component to obtain the final prediction. Extensive experiments on benchmark time-series forecasting datasets demonstrate that TDformer achieves state-of-the-art performance against existing attention-based models.

Boundary conditions (BCs) are important groups of physics-enforced constraints that are necessary for solutions of Partial Differential Equations (PDEs) to satisfy at specific spatial locations. These constraints carry important physical meaning, and guarantee the existence and the uniqueness of the PDE solution. Current neural-network based approaches that aim to solve PDEs rely only on training data to help the model learn BCs implicitly. There is no guarantee of BC satisfaction by these models during evaluation. In this work, we propose Boundary enforcing Operator Network (BOON) that enables the BC satisfaction of neural operators by making structural changes to the operator kernel. We provide our refinement procedure, and demonstrate the satisfaction of physics-based BCs, e.g. Dirichlet, Neumann, and periodic by the solutions obtained by BOON. Numerical experiments based on multiple PDEs with a wide variety of applications indicate that the proposed approach ensures satisfaction of BCs, and leads to more accurate solutions over the entire domain. The proposed correction method exhibits a (2X-20X) improvement over a given operator model in relative $L^2$ error (0.000084 relative $L^2$ error for Burgers' equation).

Training a neural network requires choosing a suitable learning rate, involving a trade-off between speed and effectiveness of convergence. While there has been considerable theoretical and empirical analysis of how large the learning rate can be, most prior work focuses only on late-stage training. In this work, we introduce the maximal initial learning rate $\eta^{\ast}$ - the largest learning rate at which a randomly initialized neural network can successfully begin training and achieve (at least) a given threshold accuracy. Using a simple approach to estimate $\eta^{\ast}$, we observe that in constant-width fully-connected ReLU networks, $\eta^{\ast}$ demonstrates different behavior to the maximum learning rate later in training. Specifically, we find that $\eta^{\ast}$ is well predicted as a power of $(\text{depth} \times \text{width})$, provided that (i) the width of the network is sufficiently large compared to the depth, and (ii) the input layer of the network is trained at a relatively small learning rate. We further analyze the relationship between $\eta^{\ast}$ and the sharpness $\lambda_{1}$ of the network at initialization, indicating that they are closely though not inversely related. We formally prove bounds for $\lambda_{1}$ in terms of $(\text{depth} \times \text{width})$ that align with our empirical results.

Multi-Scale and U-shaped Networks are widely used in various image restoration problems, including deblurring. Keeping in mind the wide range of applications, we present a comparison of these architectures and their effects on image deblurring. We also introduce a new block called as NFResblock. It consists of a Fast Fourier Transformation layer and a series of modified Non-Linear Activation Free Blocks. Based on these architectures and additions, we introduce NFResnet and NFResnet+, which are modified multi-scale and U-Net architectures, respectively. We also use three different loss functions to train these architectures: Charbonnier Loss, Edge Loss, and Frequency Reconstruction Loss. Extensive experiments on the Deep Video Deblurring dataset, along with ablation studies for each component, have been presented in this paper. The proposed architectures achieve a considerable increase in Peak Signal to Noise (PSNR) ratio and Structural Similarity Index (SSIM) value.