Vision transformers have significantly advanced the field of computer vision, offering robust modeling capabilities and global receptive field. However, their high computational demands limit their applicability in processing long sequences. To tackle this issue, State Space Models (SSMs) have gained prominence in vision tasks as they offer linear computational complexity. Recently, State Space Duality (SSD), an improved variant of SSMs, was introduced in Mamba2 to enhance model performance and efficiency. However, the inherent causal nature of SSD/SSMs restricts their applications in non-causal vision tasks. To address this limitation, we introduce Visual State Space Duality (VSSD) model, which has a non-causal format of SSD. Specifically, we propose to discard the magnitude of interactions between the hidden state and tokens while preserving their relative weights, which relieves the dependencies of token contribution on previous tokens. Together with the involvement of multi-scan strategies, we show that the scanning results can be integrated to achieve non-causality, which not only improves the performance of SSD in vision tasks but also enhances its efficiency. We conduct extensive experiments on various benchmarks including image classification, detection, and segmentation, where VSSD surpasses existing state-of-the-art SSM-based models. Code and weights are available at \url{https://github.com/YuHengsss/VSSD}.

Few-shot fine-grained human (FGH) action recognition is crucial in the context of human-robot interaction within open-set real-world environments. Existing works mainly focus on features extracted from RGB frames. However, their performances are drastically impacted in challenging scenarios, such as high-dynamic or low lighting conditions. Event cameras can independently and sparsely capture brightness changes in a scene at microsecond resolution and high dynamic range, which offer a promising solution. However, the modality differences between events and RGB frames, and the lack of paired fine-grained data hinder the development of event-based FGH action recognition. Therefore, in this paper, we introduce the first Event Camera Fine-grained Human Action (E-FAction) dataset. This dataset comprises 3304 paired ‘event stream and RGB sequence’, covering 15 coarse action classes and 128 fine-grained actions. Then, we develop a versatile event feature extractor. Considering the spatial sparsity of event stream, we design two modules to mine the temporal motion and semantic features under the guidance of paired RGB frames, facilitating robust weight initialization for the feature extractor in few-shot FGH action recognition. We conduct extensive experiments on both published and our built synthetic and real datasets, and consistently achieve state-of-the-art performance compared to existing baselines. Code and dataset will be available at link.

Despite the significant achievements of Vision Transformers (ViTs) in various vision tasks, they are constrained by the quadratic complexity. Recently, State Space Models (SSMs) have garnered widespread attention due to their global receptive field and linear complexity with respect to the input length, demonstrating substantial potential across fields including natural language processing and computer vision. To improve the performance of SSMs in vision tasks, a multi-scan strategy is widely adopted, which leads to significant redundancy of SSMs. For a better trade-off between efficiency and performance, we analyze the underlying reasons behind the success of the multi-scan strategy, where long-range dependency plays an important role. Based on the analysis, we introduce Multi-Scale Vision Mamba (MSVMamba) to preserve the superiority of SSMs in vision tasks with limited parameters. It employs a multi-scale 2D scanning technique on both original and downsampled feature maps, which not only benefits long-range dependency learning but also reduces computational costs. Additionally, we integrate a Convolutional Feed-Forward Network (ConvFFN) to address the lack of channel mixing. Our experiments demonstrate that MSVMamba is highly competitive, with the MSVMamba-Tiny model achieving 83.0% top-1 accuracy on ImageNet, 46.9% box mAP, and 42.5% instance mAP with the Mask R-CNN framework, 1x training schedule on COCO, and 47.9% mIoU with single-scale testing on ADE20K. Code is available at https://github.com/YuHengsss/MSVMamba.

Few-shot action recognition in videos is challenging as the lack of supervision makes it extremely difficult to generalize well to unseen actions. To address this challenge, we propose a simple yet effective method, called knowledge prompting, which leverages commonsense knowledge of actions from external resources to prompt-tune a powerful pre-trained vision-language model for few-shot classification. To that end, we first collect a large-scale corpus of language descriptions of actions, defined as text proposals, to build an action knowledge base. The collection of text proposals is done by filling in a handcraft sentence template with an external action-related corpus or by extracting action-related phrases from captions of Web instruction videos. Next, we feed these text proposals to a pre-trained vision-language model along with video frames to generate matching scores of the proposals for each frame, and the scores can be treated as action semantics with strong generalization. Finally, we design a lightweight temporal modeling network to capture the temporal evolution of action semantics for classification. Extensive experiments on six benchmark datasets demonstrate that our method generally achieves state-of-the-art performance while reducing the training computational cost to 0.1% of the existing methods. Code is available athttps://github.com/OldStone0124/Knowledge-Prompting-for-FSAR.

Video object detection (VID) is challenging because of the high variation of object appearance as well as the diverse deterioration in some frames. On the positive side, the detection in a certain frame of a video, compared with that in a still image, can draw support from other frames. Hence, how to aggregate features across different frames is pivotal to VID problem. Most of existing aggregation algorithms are customized for two-stage detectors. However, these detectors are usually computationally expensive due to their two-stage nature. This work proposes a simple yet effective strategy to address the above concerns, which costs marginal overheads with significant gains in accuracy. Concretely, different from traditional two-stage pipeline, we select important regions after the one-stage detection to avoid processing massive low-quality candidates. Besides, we evaluate the relationship between a target frame and reference frames to guide the aggregation. We conduct extensive experiments and ablation studies to verify the efficacy of our design, and reveal its superiority over other state-of-the-art VID approaches in both effectiveness and efficiency. Our YOLOX-based model can achieve promising performance (e.g., 87.5% AP50 at over 30 FPS on the ImageNet VID dataset on a single 2080Ti GPU), making it attractive for large-scale or real-time applications. The implementation is simple, we have made the demo codes and models available at https://github.com/YuHengsss/YOLOV.