Efficient video recognition is a hot-spot research topic with the explosive growth of multimedia data on the Internet and mobile devices. Most existing methods select the salient frames without awareness of the class-specific saliency scores, which neglect the implicit association between the saliency of frames and its belonging category. To alleviate this issue, we devise a novel Temporal Saliency Query (TSQ) mechanism, which introduces class-specific information to provide fine-grained cues for saliency measurement. Specifically, we model the class-specific saliency measuring process as a query-response task. For each category, the common pattern of it is employed as a query and the most salient frames are responded to it. Then, the calculated similarities are adopted as the frame saliency scores. To achieve it, we propose a Temporal Saliency Query Network (TSQNet) that includes two instantiations of the TSQ mechanism based on visual appearance similarities and textual event-object relations. Afterward, cross-modality interactions are imposed to promote the information exchange between them. Finally, we use the class-specific saliencies of the most confident categories generated by two modalities to perform the selection of salient frames. Extensive experiments demonstrate the effectiveness of our method by achieving state-of-the-art results on ActivityNet, FCVID and Mini-Kinetics datasets. Our project page is at https://lawrencexia2008.github.io/projects/tsqnet .
Due to the costliness of labelled data in real-world applications, semi-supervised object detectors, underpinned by pseudo labelling, are appealing. However, handling confusing samples is nontrivial: discarding valuable confusing samples would compromise the model generalisation while using them for training would exacerbate the confirmation bias issue caused by inevitable mislabelling. To solve this problem, this paper proposes to use confusing samples proactively without label correction. Specifically, a virtual category (VC) is assigned to each confusing sample such that they can safely contribute to the model optimisation even without a concrete label. It is attributed to specifying the embedding distance between the training sample and the virtual category as the lower bound of the inter-class distance. Moreover, we also modify the localisation loss to allow high-quality boundaries for location regression. Extensive experiments demonstrate that the proposed VC learning significantly surpasses the state-of-the-art, especially with small amounts of available labels.
The well-designed structures in neural networks reflect the prior knowledge incorporated into the models. However, though different models have various priors, we are used to training them with model-agnostic optimizers (e.g., SGD). In this paper, we propose a novel paradigm of incorporating model-specific prior knowledge into optimizers and using them to train generic (simple) models. As an implementation, we propose a novel methodology to add prior knowledge by modifying the gradients according to a set of model-specific hyper-parameters, which is referred to as Gradient Re-parameterization, and the optimizers are named RepOptimizers. For the extreme simplicity of model structure, we focus on a VGG-style plain model and showcase that such a simple model trained with a RepOptimizer, which is referred to as RepOpt-VGG, performs on par with the recent well-designed models. From a practical perspective, RepOpt-VGG is a favorable base model because of its simple structure, high inference speed and training efficiency. Compared to Structural Re-parameterization, which adds priors into models via constructing extra training-time structures, RepOptimizers require no extra forward/backward computations and solve the problem of quantization. The code and models are publicly available at https://github.com/DingXiaoH/RepOptimizers.
We revisit large kernel design in modern convolutional neural networks (CNNs). Inspired by recent advances in vision transformers (ViTs), in this paper, we demonstrate that using a few large convolutional kernels instead of a stack of small kernels could be a more powerful paradigm. We suggested five guidelines, e.g., applying re-parameterized large depth-wise convolutions, to design efficient high-performance large-kernel CNNs. Following the guidelines, we propose RepLKNet, a pure CNN architecture whose kernel size is as large as 31x31, in contrast to commonly used 3x3. RepLKNet greatly closes the performance gap between CNNs and ViTs, e.g., achieving comparable or superior results than Swin Transformer on ImageNet and a few typical downstream tasks, with lower latency. RepLKNet also shows nice scalability to big data and large models, obtaining 87.8% top-1 accuracy on ImageNet and 56.0% mIoU on ADE20K, which is very competitive among the state-of-the-arts with similar model sizes. Our study further reveals that, in contrast to small-kernel CNNs, large-kernel CNNs have much larger effective receptive fields and higher shape bias rather than texture bias. Code & models at https://github.com/megvii-research/RepLKNet.
Zero-shot learning (ZSL) aims to learn models that can recognize unseen image semantics based on the training of data with seen semantics. Recent studies either leverage the global image features or mine discriminative local patch features to associate the extracted visual features to the semantic attributes. However, due to the lack of the necessary top-down guidance and semantic alignment for ensuring the model attending to the real attribute-correlation regions, these methods still encounter a significant semantic gap between the visual modality and the attribute modality, which makes their prediction on unseen semantics unreliable. To solve this problem, this paper establishes a novel transformer encoder-decoder model, called hybrid routing transformer (HRT). In HRT encoder, we embed an active attention, which is constructed by both the bottom-up and the top-down dynamic routing pathways to generate the attribute-aligned visual feature. While in HRT decoder, we use static routing to calculate the correlation among the attribute-aligned visual features, the corresponding attribute semantics, and the class attribute vectors to generate the final class label predictions. This design makes the presented transformer model a hybrid of 1) top-down and bottom-up attention pathways and 2) dynamic and static routing pathways. Comprehensive experiments on three widely-used benchmark datasets, namely CUB, SUN, and AWA2, are conducted. The obtained experimental results demonstrate the effectiveness of the proposed method.
Visible-infrared person re-identification (VI-ReID) has been challenging due to the existence of large discrepancies between visible and infrared modalities. Most pioneering approaches reduce intra-class variations and inter-modality discrepancies by learning modality-shared and ID-related features. However, an explicit modality-shared cue, i.e., body keypoints, has not been fully exploited in VI-ReID. Additionally, existing feature learning paradigms imposed constraints on either global features or partitioned feature stripes, which neglect the prediction consistency of global and part features. To address the above problems, we exploit Pose Estimation as an auxiliary learning task to assist the VI-ReID task in an end-to-end framework. By jointly training these two tasks in a mutually beneficial manner, our model learns higher quality modality-shared and ID-related features. On top of it, the learnings of global features and local features are seamlessly synchronized by Hierarchical Feature Constraint (HFC), where the former supervises the latter using the knowledge distillation strategy. Experimental results on two benchmark VI-ReID datasets show that the proposed method consistently improves state-of-the-art methods by significant margins. Specifically, our method achieves nearly 20$\%$ mAP improvements against the state-of-the-art method on the RegDB dataset. Our intriguing findings highlight the usage of auxiliary task learning in VI-ReID.
Recent advances in machine learning and prevalence of digital medical images have opened up an opportunity to address the challenging brain tumor segmentation (BTS) task by using deep convolutional neural networks. However, different from the RGB image data that are very widespread, the medical image data used in brain tumor segmentation are relatively scarce in terms of the data scale but contain the richer information in terms of the modality property. To this end, this paper proposes a novel cross-modality deep feature learning framework to segment brain tumors from the multi-modality MRI data. The core idea is to mine rich patterns across the multi-modality data to make up for the insufficient data scale. The proposed cross-modality deep feature learning framework consists of two learning processes: the cross-modality feature transition (CMFT) process and the cross-modality feature fusion (CMFF) process, which aims at learning rich feature representations by transiting knowledge across different modality data and fusing knowledge from different modality data, respectively. Comprehensive experiments are conducted on the BraTS benchmarks, which show that the proposed cross-modality deep feature learning framework can effectively improve the brain tumor segmentation performance when compared with the baseline methods and state-of-the-art methods.
Compared to convolutional layers, fully-connected (FC) layers are better at modeling the long-range dependencies but worse at capturing the local patterns, hence usually less favored for image recognition. In this paper, we propose a methodology, Locality Injection, to incorporate local priors into an FC layer via merging the trained parameters of a parallel conv kernel into the FC kernel. Locality Injection can be viewed as a novel Structural Re-parameterization method since it equivalently converts the structures via transforming the parameters. Based on that, we propose a multi-layer-perceptron (MLP) block named RepMLP Block, which uses three FC layers to extract features, and a novel architecture named RepMLPNet. The hierarchical design distinguishes RepMLPNet from the other concurrently proposed vision MLPs. As it produces feature maps of different levels, it qualifies as a backbone model for downstream tasks like semantic segmentation. Our results reveal that 1) Locality Injection is a general methodology for MLP models; 2) RepMLPNet has favorable accuracy-efficiency trade-off compared to the other MLPs; 3) RepMLPNet is the first MLP that seamlessly transfer to Cityscapes semantic segmentation. The code and models are available at https://github.com/DingXiaoH/RepMLP.
While recent deep deblurring algorithms have achieved remarkable progress, most existing methods focus on the global deblurring problem, where the image blur mostly arises from severe camera shake. We argue that the local blur, which is mostly derived from moving objects with a relatively static background, is prevalent but remains under-explored. In this paper, we first lay the data foundation for local deblurring by constructing, for the first time, a LOcal-DEblur (LODE) dataset consisting of 3,700 real-world captured locally blurred images and their corresponding ground-truth. Then, we propose a novel framework, termed BLur-Aware DEblurring network (BladeNet), which contains three components: the Local Blur Synthesis module generates locally blurred training pairs, the Local Blur Perception module automatically captures the locally blurred region and the Blur-guided Spatial Attention module guides the deblurring network with spatial attention. This framework is flexible such that it can be combined with many existing SotA algorithms. We carry out extensive experiments on REDS and LODE datasets showing that BladeNet improves PSNR by 2.5dB over SotAs for local deblurring while keeping comparable performance for global deblurring. We will publish the dataset and codes.