By adopting popular pixel-wise loss, existing methods for defocus deblurring heavily rely on well aligned training image pairs. Although training pairs of ground-truth and blurry images are carefully collected, e.g., DPDD dataset, misalignment is inevitable between training pairs, making existing methods possibly suffer from deformation artifacts. In this paper, we propose a joint deblurring and reblurring learning (JDRL) framework for single image defocus deblurring with misaligned training pairs. Generally, JDRL consists of a deblurring module and a spatially invariant reblurring module, by which deblurred result can be adaptively supervised by ground-truth image to recover sharp textures while maintaining spatial consistency with the blurry image. First, in the deblurring module, a bi-directional optical flow-based deformation is introduced to tolerate spatial misalignment between deblurred and ground-truth images. Second, in the reblurring module, deblurred result is reblurred to be spatially aligned with blurry image, by predicting a set of isotropic blur kernels and weighting maps. Moreover, we establish a new single image defocus deblurring (SDD) dataset, further validating our JDRL and also benefiting future research. Our JDRL can be applied to boost defocus deblurring networks in terms of both quantitative metrics and visual quality on DPDD, RealDOF and our SDD datasets.

最近，基于深度学习的图像降级方法在测试数据上具有与训练集相同的测试数据的有希望的性能，在该数据中，已经学习了基于合成或收集的现实世界训练数据的各种denoising模型。但是，在处理真实世界的嘈杂图像时，Denoising的性能仍然受到限制。在本文中，我们提出了一种简单而有效的贝叶斯深集合（BDE）方法，用于真实世界图像denoising，其中可以融合使用各种训练数据设置进行预训练的几位代表性的深层Denoiser，以提高稳健性。 BDE的基础是，现实世界的图像噪声高度取决于信号依赖性，并且在现实世界中的嘈杂图像中的异质噪声可以由不同的Deoisiser分别处理。特别是，我们将受过良好训练的CBDNET，NBNET，HINET，UFORFORMER和GMSNET进入Denoiser池，并采用U-NET来预测Pixel的加权图以融合这些DeOisiser。引入了贝叶斯深度学习策略，而不是仅仅学习像素的加权地图，而是为了预测加权不确定性和加权图，可以通过该策略来建模预测差异，以改善现实世界中的嘈杂图像的鲁棒性。广泛的实验表明，可以通过融合现有的DINOISER而不是训练一个以昂贵的成本来训练一个大的Denoiser来更好地消除现实世界的噪音。在DND数据集上，我们的BDE实现了 +0.28〜dB PSNR的增益，而不是最先进的denoising方法。此外，我们注意到，在应用于现实世界嘈杂的图像时，基于不同高斯噪声水平的BDE Denoiser优于最先进的CBDNET。此外，我们的BDE可以扩展到其他图像恢复任务，并在基准数据集上获得 +0.30dB， +0.18dB和 +0.12dB PSNR的收益，以分别用于图像去除图像，图像降低和单个图像超级分辨率。

Previous knowledge distillation (KD) methods for object detection mostly focus on feature imitation instead of mimicking the prediction logits due to its inefficiency in distilling the localization information. In this paper, we investigate whether logit mimicking always lags behind feature imitation. Towards this goal, we first present a novel localization distillation (LD) method which can efficiently transfer the localization knowledge from the teacher to the student. Second, we introduce the concept of valuable localization region that can aid to selectively distill the classification and localization knowledge for a certain region. Combining these two new components, for the first time, we show that logit mimicking can outperform feature imitation and the absence of localization distillation is a critical reason for why logit mimicking underperforms for years. The thorough studies exhibit the great potential of logit mimicking that can significantly alleviate the localization ambiguity, learn robust feature representation, and ease the training difficulty in the early stage. We also provide the theoretical connection between the proposed LD and the classification KD, that they share the equivalent optimization effect. Our distillation scheme is simple as well as effective and can be easily applied to both dense horizontal object detectors and rotated object detectors. Extensive experiments on the MS COCO, PASCAL VOC, and DOTA benchmarks demonstrate that our method can achieve considerable AP improvement without any sacrifice on the inference speed. Our source code and pretrained models are publicly available at https://github.com/HikariTJU/LD.

知识蒸馏（KD）目睹了其在物体检测中学习紧凑型号的强大能力。以前的KD方法用于对象检测主要是侧重于模仿仿地区内的深度特征，而不是模仿分类登录，而不是蒸馏定位信息的低效率。在本文中，通过重新制定本地化的知识蒸馏过程，我们提出了一种新的本地化蒸馏（LD）方法，可以有效地将老师的本地化知识转移给学生。此外，我们还启发式介绍了有价值的本地化区域的概念，可以帮助选择性地蒸馏某个地区的语义和本地化知识。第一次结合这两个新组件，我们显示Logit Mimicing可以优于特征模仿和本地化知识蒸馏比蒸馏对象探测器的语义知识更为重要和有效。我们的蒸馏方案简单，有效，可以很容易地应用于不同的致密物体探测器。实验表明，我们的LD可以将GFOCal-Reset-50的AP得分提升，单一规模的1 $ \ Times $培训计划从Coco基准测试中的40.1到42.1，没有任何牺牲品推断速度。我们的源代码和培训的型号在https://github.com/hikaritju/ld公开提供

Bounding box regression is the crucial step in object detection. In existing methods, while n-norm loss is widely adopted for bounding box regression, it is not tailored to the evaluation metric, i.e., Intersection over Union (IoU). Recently, IoU loss and generalized IoU (GIoU) loss have been proposed to benefit the IoU metric, but still suffer from the problems of slow convergence and inaccurate regression. In this paper, we propose a Distance-IoU (DIoU) loss by incorporating the normalized distance between the predicted box and the target box, which converges much faster in training than IoU and GIoU losses. Furthermore, this paper summarizes three geometric factors in bounding box regression, i.e., overlap area, central point distance and aspect ratio, based on which a Complete IoU (CIoU) loss is proposed, thereby leading to faster convergence and better performance. By incorporating DIoU and CIoU losses into state-of-the-art object detection algorithms, e.g., YOLO v3, SSD and Faster R-CNN, we achieve notable performance gains in terms of not only IoU metric but also GIoU metric. Moreover, DIoU can be easily adopted into non-maximum suppression (NMS) to act as the criterion, further boosting performance improvement. The source code and trained models are available at https://github.com/Zzh-tju/DIoU.

Masked image modeling (MIM) performs strongly in pre-training large vision Transformers (ViTs). However, small models that are critical for real-world applications cannot or only marginally benefit from this pre-training approach. In this paper, we explore distillation techniques to transfer the success of large MIM-based pre-trained models to smaller ones. We systematically study different options in the distillation framework, including distilling targets, losses, input, network regularization, sequential distillation, etc, revealing that: 1) Distilling token relations is more effective than CLS token- and feature-based distillation; 2) An intermediate layer of the teacher network as target perform better than that using the last layer when the depth of the student mismatches that of the teacher; 3) Weak regularization is preferred; etc. With these findings, we achieve significant fine-tuning accuracy improvements over the scratch MIM pre-training on ImageNet-1K classification, using all the ViT-Tiny, ViT-Small, and ViT-base models, with +4.2%/+2.4%/+1.4% gains, respectively. Our TinyMIM model of base size achieves 52.2 mIoU in AE20K semantic segmentation, which is +4.1 higher than the MAE baseline. Our TinyMIM model of tiny size achieves 79.6% top-1 accuracy on ImageNet-1K image classification, which sets a new record for small vision models of the same size and computation budget. This strong performance suggests an alternative way for developing small vision Transformer models, that is, by exploring better training methods rather than introducing inductive biases into architectures as in most previous works. Code is available at https://github.com/OliverRensu/TinyMIM.

This paper presents a practical global optimization algorithm for the K-center clustering problem, which aims to select K samples as the cluster centers to minimize the maximum within-cluster distance. This algorithm is based on a reduced-space branch and bound scheme and guarantees convergence to the global optimum in a finite number of steps by only branching on the regions of centers. To improve efficiency, we have designed a two-stage decomposable lower bound, the solution of which can be derived in a closed form. In addition, we also propose several acceleration techniques to narrow down the region of centers, including bounds tightening, sample reduction, and parallelization. Extensive studies on synthetic and real-world datasets have demonstrated that our algorithm can solve the K-center problems to global optimal within 4 hours for ten million samples in the serial mode and one billion samples in the parallel mode. Moreover, compared with the state-of-the-art heuristic methods, the global optimum obtained by our algorithm can averagely reduce the objective function by 25.8% on all the synthetic and real-world datasets.

Score-based diffusion models have captured widespread attention and funded fast progress of recent vision generative tasks. In this paper, we focus on diffusion model backbone which has been much neglected before. We systematically explore vision Transformers as diffusion learners for various generative tasks. With our improvements the performance of vanilla ViT-based backbone (IU-ViT) is boosted to be on par with traditional U-Net-based methods. We further provide a hypothesis on the implication of disentangling the generative backbone as an encoder-decoder structure and show proof-of-concept experiments verifying the effectiveness of a stronger encoder for generative tasks with ASymmetriC ENcoder Decoder (ASCEND). Our improvements achieve competitive results on CIFAR-10, CelebA, LSUN, CUB Bird and large-resolution text-to-image tasks. To the best of our knowledge, we are the first to successfully train a single diffusion model on text-to-image task beyond 64x64 resolution. We hope this will motivate people to rethink the modeling choices and the training pipelines for diffusion-based generative models.

Deep learning-based methods have achieved significant performance for image defogging. However, existing methods are mainly developed for land scenes and perform poorly when dealing with overwater foggy images, since overwater scenes typically contain large expanses of sky and water. In this work, we propose a Prior map Guided CycleGAN (PG-CycleGAN) for defogging of images with overwater scenes. To promote the recovery of the objects on water in the image, two loss functions are exploited for the network where a prior map is designed to invert the dark channel and the min-max normalization is used to suppress the sky and emphasize objects. However, due to the unpaired training set, the network may learn an under-constrained domain mapping from foggy to fog-free image, leading to artifacts and loss of details. Thus, we propose an intuitive Upscaling Inception Module (UIM) and a Long-range Residual Coarse-to-fine framework (LRC) to mitigate this issue. Extensive experiments on qualitative and quantitative comparisons demonstrate that the proposed method outperforms the state-of-the-art supervised, semi-supervised, and unsupervised defogging approaches.

Conversational recommender systems (CRSs) often utilize external knowledge graphs (KGs) to introduce rich semantic information and recommend relevant items through natural language dialogues. However, original KGs employed in existing CRSs are often incomplete and sparse, which limits the reasoning capability in recommendation. Moreover, only few of existing studies exploit the dialogue context to dynamically refine knowledge from KGs for better recommendation. To address the above issues, we propose the Variational Reasoning over Incomplete KGs Conversational Recommender (VRICR). Our key idea is to incorporate the large dialogue corpus naturally accompanied with CRSs to enhance the incomplete KGs; and perform dynamic knowledge reasoning conditioned on the dialogue context. Specifically, we denote the dialogue-specific subgraphs of KGs as latent variables with categorical priors for adaptive knowledge graphs refactor. We propose a variational Bayesian method to approximate posterior distributions over dialogue-specific subgraphs, which not only leverages the dialogue corpus for restructuring missing entity relations but also dynamically selects knowledge based on the dialogue context. Finally, we infuse the dialogue-specific subgraphs to decode the recommendation and responses. We conduct experiments on two benchmark CRSs datasets. Experimental results confirm the effectiveness of our proposed method.