In this paper, we present the first neural video codec that can compete with the latest coding standard H.266/VVC in terms of sRGB PSNR on UVG dataset for the low-latency mode. Existing neural hybrid video coding approaches rely on optical flow or Gaussian-scale flow for prediction, which cannot support fine-grained adaptation to diverse motion content. Towards more content-adaptive prediction, we propose a novel cross-scale prediction module that achieves more effective motion compensation. Specifically, on the one hand, we produce a reference feature pyramid as prediction sources, then transmit cross-scale flows that leverage the feature scale to control the precision of prediction. On the other hand, we introduce the mechanism of weighted prediction into the scenario of prediction with a single reference frame, where cross-scale weight maps are transmitted to synthesize a fine prediction result. In addition to the cross-scale prediction module, we further propose a multi-stage quantization strategy, which improves the rate-distortion performance with no extra computational penalty during inference. We show the encouraging performance of our efficient neural video codec (ENVC) on several common benchmark datasets and analyze in detail the effectiveness of every important component.
Data augmentation (DA) has been widely investigated to facilitate model optimization in many tasks. However, in most cases, data augmentation is randomly performed for each training sample with a certain probability, which might incur content destruction and visual ambiguities. To eliminate this, in this paper, we propose an effective approach, dubbed SelectAugment, to select samples to be augmented in a deterministic and online manner based on the sample contents and the network training status. Specifically, in each batch, we first determine the augmentation ratio, and then decide whether to augment each training sample under this ratio. We model this process as a two-step Markov decision process and adopt Hierarchical Reinforcement Learning (HRL) to learn the augmentation policy. In this way, the negative effects of the randomness in selecting samples to augment can be effectively alleviated and the effectiveness of DA is improved. Extensive experiments demonstrate that our proposed SelectAugment can be adapted upon numerous commonly used DA methods, e.g., Mixup, Cutmix, AutoAugment, etc, and improve their performance on multiple benchmark datasets of image classification and fine-grained image recognition.
Confounders in deep learning are in general detrimental to model's generalization where they infiltrate feature representations. Therefore, learning causal features that are free of interference from confounders is important. Most previous causal learning based approaches employ back-door criterion to mitigate the adverse effect of certain specific confounder, which require the explicit identification of confounder. However, in real scenarios, confounders are typically diverse and difficult to be identified. In this paper, we propose a novel Confounder Identification-free Causal Visual Feature Learning (CICF) method, which obviates the need for identifying confounders. CICF models the interventions among different samples based on front-door criterion, and then approximates the global-scope intervening effect upon the instance-level interventions from the perspective of optimization. In this way, we aim to find a reliable optimization direction, which avoids the intervening effects of confounders, to learn causal features. Furthermore, we uncover the relation between CICF and the popular meta-learning strategy MAML, and provide an interpretation of why MAML works from the theoretical perspective of causal learning for the first time. Thanks to the effective learning of causal features, our CICF enables models to have superior generalization capability. Extensive experiments on domain generalization benchmark datasets demonstrate the effectiveness of our CICF, which achieves the state-of-the-art performance.
Collecting large clean-distorted training image pairs in real world is non-trivial, which seriously limits the practical applications of these supervised learning based image restoration (IR) methods. Previous works attempt to address this problem by leveraging unsupervised learning technologies to alleviate the dependency for paired training samples. However, these methods typically suffer from unsatisfactory textures synthesis due to the lack of clean image supervision. Compared with purely unsupervised solution, the under-explored scheme with Few-Shot clean images (FS-IR) is more feasible to tackle this challenging real Image Restoration task. In this paper, we are the first to investigate the few-shot real image restoration and propose a Distortion-Relation guided Transfer Learning (termed as DRTL) framework. DRTL assigns a knowledge graph to capture the distortion relation between auxiliary tasks (i.e., synthetic distortions) and target tasks (i.e., real distortions with few images), and then adopt a gradient weighting strategy to guide the knowledge transfer from auxiliary task to target task. In this way, DRTL could quickly learn the most relevant knowledge from the prior distortions for target distortion. We instantiate DRTL integrated with pre-training and meta-learning pipelines as an embodiment to realize a distortion-relation aware FS-IR. Extensive experiments on multiple benchmarks demonstrate the effectiveness of DRTL on few-shot real image restoration.
Unsupervised Person Re-identification (U-ReID) with pseudo labeling recently reaches a competitive performance compared to fully-supervised ReID methods based on modern clustering algorithms. However, such clustering-based scheme becomes computationally prohibitive for large-scale datasets. How to efficiently leverage endless unlabeled data with limited computing resources for better U-ReID is under-explored. In this paper, we make the first attempt to the large-scale U-ReID and propose a "small data for big task" paradigm dubbed Meta Clustering Learning (MCL). MCL only pseudo-labels a subset of the entire unlabeled data via clustering to save computing for the first-phase training. After that, the learned cluster centroids, termed as meta-prototypes in our MCL, are regarded as a proxy annotator to softly annotate the rest unlabeled data for further polishing the model. To alleviate the potential noisy labeling issue in the polishment phase, we enforce two well-designed loss constraints to promise intra-identity consistency and inter-identity strong correlation. For multiple widely-used U-ReID benchmarks, our method significantly saves computational cost while achieving a comparable or even better performance compared to prior works.
Despite the recent progress in light field super-resolution (LFSR) achieved by convolutional neural networks, the correlation information of light field (LF) images has not been sufficiently studied and exploited due to the complexity of 4D LF data. To cope with such high-dimensional LF data, most of the existing LFSR methods resorted to decomposing it into lower dimensions and subsequently performing optimization on the decomposed sub-spaces. However, these methods are inherently limited as they neglected the characteristics of the decomposition operations and only utilized a limited set of LF sub-spaces ending up failing to comprehensively extract spatio-angular features and leading to a performance bottleneck. To overcome these limitations, in this paper, we thoroughly discover the potentials of LF decomposition and propose a novel concept of decomposition kernels. In particular, we systematically unify the decomposition operations of various sub-spaces into a series of such decomposition kernels, which are incorporated into our proposed Decomposition Kernel Network (DKNet) for comprehensive spatio-angular feature extraction. The proposed DKNet is experimentally verified to achieve substantial improvements by 1.35 dB, 0.83 dB, and 1.80 dB PSNR in 2x, 3x and 4x LFSR scales, respectively, when compared with the state-of-the-art methods. To further improve DKNet in producing more visually pleasing LFSR results, based on the VGG network, we propose a LFVGG loss to guide the Texture-Enhanced DKNet (TE-DKNet) to generate rich authentic textures and enhance LF images' visual quality significantly. We also propose an indirect evaluation metric by taking advantage of LF material recognition to objectively assess the perceptual enhancement brought by the LFVGG loss.
Learned video compression methods have demonstrated great promise in catching up with traditional video codecs in their rate-distortion (R-D) performance. However, existing learned video compression schemes are limited by the binding of the prediction mode and the fixed network framework. They are unable to support various inter prediction modes and thus inapplicable for various scenarios. In this paper, to break this limitation, we propose a versatile learned video compression (VLVC) framework that uses one model to support all possible prediction modes. Specifically, to realize versatile compression, we first build a motion compensation module that applies multiple 3D motion vector fields (i.e., voxel flows) for weighted trilinear warping in spatial-temporal space. The voxel flows convey the information of temporal reference position that helps to decouple inter prediction modes away from framework designing. Secondly, in case of multiple-reference-frame prediction, we apply a flow prediction module to predict accurate motion trajectories with a unified polynomial function. We show that the flow prediction module can largely reduce the transmission cost of voxel flows. Experimental results demonstrate that our proposed VLVC not only supports versatile compression in various settings but also achieves comparable R-D performance with the latest VVC standard in terms of MS-SSIM.
Unsupervised domain adaptive classification intends to improve theclassification performance on unlabeled target domain. To alleviate the adverse effect of domain shift, many approaches align the source and target domains in the feature space. However, a feature is usually taken as a whole for alignment without explicitly making domain alignment proactively serve the classification task, leading to sub-optimal solution. What sub-feature should be aligned for better adaptation is under-explored. In this paper, we propose an effective Task-oriented Alignment (ToAlign) for unsupervised domain adaptation (UDA). We study what features should be aligned across domains and propose to make the domain alignment proactively serve classification by performing feature decomposition and alignment under the guidance of the prior knowledge induced from the classification taskitself. Particularly, we explicitly decompose a feature in the source domain intoa task-related/discriminative feature that should be aligned, and a task-irrelevant feature that should be avoided/ignored, based on the classification meta-knowledge. Extensive experimental results on various benchmarks (e.g., Office-Home, Visda-2017, and DomainNet) under different domain adaptation settings demonstrate theeffectiveness of ToAlign which helps achieve the state-of-the-art performance.
Learning good feature representations is important for deep reinforcement learning (RL). However, with limited experience, RL often suffers from data inefficiency for training. For un-experienced or less-experienced trajectories (i.e., state-action sequences), the lack of data limits the use of them for better feature learning. In this work, we propose a novel method, dubbed PlayVirtual, which augments cycle-consistent virtual trajectories to enhance the data efficiency for RL feature representation learning. Specifically, PlayVirtual predicts future states based on the current state and action by a dynamics model and then predicts the previous states by a backward dynamics model, which forms a trajectory cycle. Based on this, we augment the actions to generate a large amount of virtual state-action trajectories. Being free of groudtruth state supervision, we enforce a trajectory to meet the cycle consistency constraint, which can significantly enhance the data efficiency. We validate the effectiveness of our designs on the Atari and DeepMind Control Suite benchmarks. Our method outperforms the current state-of-the-art methods by a large margin on both benchmarks.
Task-driven semantic video/image coding has drawn considerable attention with the development of intelligent media applications, such as license plate detection, face detection, and medical diagnosis, which focuses on maintaining the semantic information of videos/images. Deep neural network (DNN)-based codecs have been studied for this purpose due to their inherent end-to-end optimization mechanism. However, the traditional hybrid coding framework cannot be optimized in an end-to-end manner, which makes task-driven semantic fidelity metric unable to be automatically integrated into the rate-distortion optimization process. Therefore, it is still attractive and challenging to implement task-driven semantic coding with the traditional hybrid coding framework, which should still be widely used in practical industry for a long time. To solve this challenge, we design semantic maps for different tasks to extract the pixelwise semantic fidelity for videos/images. Instead of directly integrating the semantic fidelity metric into traditional hybrid coding framework, we implement task-driven semantic coding by implementing semantic bit allocation based on reinforcement learning (RL). We formulate the semantic bit allocation problem as a Markov decision process (MDP) and utilize one RL agent to automatically determine the quantization parameters (QPs) for different coding units (CUs) according to the task-driven semantic fidelity metric. Extensive experiments on different tasks, such as classification, detection and segmentation, have demonstrated the superior performance of our approach by achieving an average bitrate saving of 34.39% to 52.62% over the High Efficiency Video Coding (H.265/HEVC) anchor under equivalent task-related semantic fidelity.