This paper is not motivated to seek innovation within the attention mechanism. Instead, it focuses on overcoming the existing trade-offs between accuracy and efficiency within the context of point cloud processing, leveraging the power of scale. Drawing inspiration from recent advances in 3D large-scale representation learning, we recognize that model performance is more influenced by scale than by intricate design. Therefore, we present Point Transformer V3 (PTv3), which prioritizes simplicity and efficiency over the accuracy of certain mechanisms that are minor to the overall performance after scaling, such as replacing the precise neighbor search by KNN with an efficient serialized neighbor mapping of point clouds organized with specific patterns. This principle enables significant scaling, expanding the receptive field from 16 to 1024 points while remaining efficient (a 3x increase in processing speed and a 10x improvement in memory efficiency compared with its predecessor, PTv2). PTv3 attains state-of-the-art results on over 20 downstream tasks that span both indoor and outdoor scenarios. Further enhanced with multi-dataset joint training, PTv3 pushes these results to a higher level.
Simpler and Stronger Baselines for All-in-One Image Restoration
School of Mathematics, Tianjin University(天津大学数学学院);School of Cybersecurity, Tianjin University(天津大学网络安全学院);School of Artificial Intelligence, Tianjin University(天津大学人工智能学院)
MC-PanDA++: Simpler, Stronger, and More Robust Domain-Adaptive Panoptic Segmentation
Faculty of Electrical Engineering and Computing, University of Zagreb(萨格勒布大学电气工程与计算学院);Faculty of Computer and Information Science, University of Ljubljana(卢布尔雅那大学计算机与信息科学学院);Fundamental AI Lab, University of Technology Nuremberg(纽伦堡工业大学基础人工智能实验室)
Faster and Simpler Siamese Network for Single Object Tracking
State Key Laboratory for Novel Software Technology, Nanjing University(南京大学计算机软件新技术国家重点实验室);Department of Computer Science and Technology, Nanjing University(南京大学计算机科学与技术系);School of Electronic Science and Engineering, Nanjing University(南京大学电子科学与工程学院)
Constructing Stronger and Faster Baselines for Skeleton-based Action Recognition
University of Chinese Academy of Sciences(中国科学院大学);Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所);Shandong University of Science and Technology(山东科技大学);Artificial Intelligence Research, Chinese Academy of Sciences(中国科学院人工智能研究院)
A Faster, Lighter and Stronger Deep Learning-Based Approach for Place Recognition
Xiamen University(厦门大学)
视觉位置识别是摄像机定位和环路闭合检测系统的重要组成部分,在计算机视觉、机器人和AR/VR等多个领域引起了广泛的兴趣。在这项工作中,我们提出了一种更快,更轻,更强大的方法,可以生成具有更少参数的模型,并且可以在推理阶段花费更少的时间。我们设计了RepVGG-lite作为我们的架构中的骨干网络,它比其他一般网络在地点识别任务中更具鉴别力。RepVGG-lite在实现更高性能的同时具有更多的速度优势。在特征提取阶段,我们只从全局描述符中提取一个尺度的块级描述符。然后,我们设计了一个可训练的特征匹配器,利用的空间关系的功能和他们的视觉外观,这是基于注意力机制。在具有挑战性的基准数据集上进行的综合实验表明,所提出的方法优于最近其他最先进的学习方法,并实现了更高的推理速度。我们的系统比Patch-NetVLAD少14倍的参数,低6.8倍的理论FLOPs,并且在特征提取和特征匹配中运行速度快21和33倍。此外,我们的方法的性能是0.5\%,比补丁NetVLAD在Recall@1。我们使用Mapillary Street Level Sequences数据集的子集对所有其他具有挑战性的条件进行实验。
Stronger, Faster and More Explainable: A Graph Convolutional Baseline for Skeleton-based Action Recognition
University of Chinese Academy of Sciences(中国科学院大学);Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所);College of Electrical Engineering and Automation, Shandong University of Science and Technology(山东科技大学电气工程与自动化学院);School of Computer Science and Technology, Anhui University(安徽大学计算机科学与技术学院)
Lumina-Next: Making Lumina-T2X Stronger and Faster with Next-DiT
Shanghai AI Laboratory(上海人工智能实验室);The Chinese University of Hong Kong(香港中文大学);Beijing University of Posts and Telecommunications(北京邮电大学);Beihang University(北京航空航天大学);Nanyang Technological University(南洋理工大学)