Real-time video generation with Diffusion Transformers is bottlenecked by the quadratic cost of 3D self-attention, especially in real-time regimes that are both few-step and autoregressive, where errors compound across time and each denoising step must carry substantially more information. In this setting, we find that prior sparse-attention approximations break down, despite showing strong results for bidirectional, many-step diffusion. Specifically, we observe that video attention is not reliably sparse, but instead combines pronounced periodic structure driven by spatiotemporal position with dynamic, sparse semantic correspondences and dense mixing, exceeding the representational capacity of even oracle top-k attention. Building on this insight, we propose Monarch-RT, a structured attention parameterization for video diffusion models that factorizes attention using Monarch matrices. Through appropriately aligned block structure and our extended tiled Monarch parameterization, we achieve high expressivity while preserving computational efficiency. We further overcome the overhead of parameterization through finetuning, with custom Triton kernels. We first validate the high efficacy of Monarch-RT over existing sparse baselines designed only for bidirectional models. We further observe that Monarch-RT attains up to 95% attention sparsity with no loss in quality when applied to the state-of-the-art model Self-Forcing, making Monarch-RT a pioneering work on highly-capable sparse attention parameterization for real-time video generation. Our optimized implementation outperforms FlashAttention-2, FlashAttention-3, and FlashAttention-4 kernels on Nvidia RTX 5090, H100, and B200 GPUs respectively, providing kernel speedups in the range of 1.4-11.8X. This enables us, for the first time, to achieve true real-time video generation with Self-Forcing at 16 FPS on a single RTX 5090.
The Chinese University of Hong Kong（香港中文大学）；Adobe（奥多比）；SmartMore（思谋）
This paper presents Video-P2P, a novel framework for real-world video editing with cross-attention control. While attention control has proven effective for image editing with pre-trained image generation models, there are currently no large-scale video generation models publicly available. Video-P2P addresses this limitation by adapting an image generation diffusion model to complete various video editing tasks. Specifically, we propose to first tune a Text-to-Set (T2S) model to complete an approximate inversion and then optimize a shared unconditional embedding to achieve accurate video inve…
VMC: Video Motion Customization using Temporal Attention Adaption for Text-to-Video Diffusion Models
Kim Jaechul Graduate School of AI（金宰澈人工智能研究生院）；Bio and Brain Engineering（生物与脑工程系）；Korea Advanced Institute of Science and Technology (KAIST)（韩国科学技术院（KAIST））
We proposed a novel trimap free video matting method based on the attention mechanism. By the nature of the problem, most existing approaches use either multiple computational expansive modules or complex algorithms to exploit temporal information fully. We designed a temporal aggregation module to compute the temporal coherence between the current frame and its two previous frames.
The Australian National University（澳大利亚国立大学）；DATA61-CSIRO（澳大利亚联邦科学与工业研究组织数据61部）；NEC Laboratories America（美国电气股份有限公司实验室）；OPPO US Research Center（OPPO美国研究中心）；Monash University（莫纳什大学）
Full attention, which generates an attention value per element of the input feature maps, has been successfully demonstrated to be beneficial in visual tasks. In this work, we propose a fully attentional network, termed {\it channel recurrent attention network}, for the task of video pedestrian retrieval. The main attention unit, \textit{channel recurrent attention}, identifies attention maps at the frame level by jointly leveraging spatial and channel patterns via a recurrent neural network. This channel recurrent attention is designed to build a global receptive field by recurrently receivin…
We propose MASTAF, a Model-Agnostic Spatio-Temporal Attention Fusion network for few-shot video classification. MASTAF takes input from a general video spatial and temporal representation,e.g., using 2D CNN, 3D CNN, and Video Transformer. Then, to make the most of such representations, we use self- and cross-attention models to highlight the critical spatio-temporal region to increase the inter-class variations and decrease the intra-class variations. Last, MASTAF applies a lightweight fusion network and a nearest neighbor classifier to classify each query video. We demonstrate that MASTAF imp…
Insight Centre for Data Analytics, Dublin City University（都柏林城市大学Insight数据分析中心）；Institut National des Télécommunications et des TIC（国家电信与信息通信技术学院）
The spherical domain representation of 360 video/image presents many challenges related to the storage, processing, transmission and rendering of omnidirectional videos (ODV). Models of human visual attention can be used so that only a single viewport is rendered at a time, which is important when developing systems that allow users to explore ODV with head mounted displays (HMD). Accordingly, researchers have proposed various saliency models for 360 video/images. This paper proposes ATSal, a novel attention based (head-eye) saliency model for 360\degree videos. The attention mechanism explici…
Tsinghua University（清华大学）；Rutgers University（罗格斯大学）；Massachusetts Institute of Technology（麻省理工学院）；Baidu IDL（百度深度学习研究院）
Recently, substantial research effort has focused on how to apply CNNs or RNNs to better extract temporal patterns from videos, so as to improve the accuracy of video classification. In this paper, however, we show that temporal information, especially longer-term patterns, may not be necessary to achieve competitive results on common video classification datasets. We investigate the potential of a purely attention based local feature integration. Accounting for the characteristics of such features in video classification, we propose a local feature integration framework based on attention clu…
Renmin University of China（中国人民大学）；University of Adelaide（阿德莱德大学）
This notebook paper presents our model in the VATEX video captioning challenge. In order to capture multi-level aspects in the video, we propose to integrate both temporal and spatial attentions for video captioning. The temporal attentive module focuses on global action movements while spatial attentive module enables to describe more fine-grained objects. Considering these two types of attentive modules are complementary, we thus fuse them via a late fusion strategy. The proposed model significantly outperforms baselines and achieves 73.4 CIDEr score on the testing set which ranks the second…
Zhejiang University（浙江大学）；University of Electronic Science and Technology of China（电子科技大学）
Open-ended video question answering aims to automatically generate the natural-language answer from referenced video contents according to the given question. Currently, most existing approaches focus on short-form video question answering with multi-modal recurrent encoder-decoder networks. Although these works have achieved promising performance, they may still be ineffectively applied to long-form video question answering due to the lack of long-range dependency modeling and the suffering from the heavy computational cost. To tackle these problems, we propose a fast Hierarchical Convolution…
The significant growth of surveillance camera networks necessitates scalable AI solutions to efficiently analyze the large amount of video data produced by these networks. As a typical analysis performed on surveillance footage, video violence detection has recently received considerable attention. The majority of research has focused on improving existing methods using supervised methods, with little, if any, attention to the semi-supervised learning approaches. In this study, a reinforcement learning model is introduced that can outperform existing models through a semi-supervised approach. …
Samsung AI Center（三星人工智能中心）；Queen Mary University London（伦敦玛丽女王大学）
This paper tackles the problem of efficient video recognition. In this area, video transformers have recently dominated the efficiency (top-1 accuracy vs FLOPs) spectrum. At the same time, there have been some attempts in the image domain which challenge the necessity of the self-attention operation within the transformer architecture, advocating the use of simpler approaches for token mixing. However, there are no results yet for the case of video recognition, where the self-attention operator has a significantly higher impact (compared to the case of images) on efficiency. To address this ga…
Suzhou Vocational Institute of Industrial Technology（苏州工业职业技术学院）；Texas Tech University（德克萨斯理工大学）；Suzhou University of Science and Technology（苏州科技大学）
In order to deal with variant-length long videos, prior works extract multi-modal features and fuse them to predict students' engagement intensity. In this paper, we present a new end-to-end method Class Attention in Video Transformer (CavT), which involves a single vector to process class embedding and to uniformly perform end-to-end learning on variant-length long videos and fixed-length short videos. Furthermore, to address the lack of sufficient samples, we propose a binary-order representatives sampling method (BorS) to add multiple video sequences of each video to augment the training se…
University of Texas at Arlington（德克萨斯大学阿灵顿分校）；Google（谷歌）；Microsoft Research（微软研究院）
We present a method to improve video description generation by modeling higher-order interactions between video frames and described concepts. By storing past visual attention in the video associated to previously generated words, the system is able to decide what to look at and describe in light of what it has already looked at and described. This enables not only more effective local attention, but tractable consideration of the video sequence while generating each word. Evaluation on the challenging and popular MSVD and Charades datasets demonstrates that the proposed architecture outperfor…
