.png){kind=icon}
返回
Monocular 3D-trajectory reconstruction using models-driven weakly supervised learning
.png){kind=icon}
OA
AI
DOI：10.1016/j.optlaseng.2023.107798.png)
.png){kind=icon}
.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Neural networks have demonstrated remarkable success in various computer vision tasks, offering benefits such as self-learning, self-organization, and self-adaptation. To maintain these advantages, a high-quality labeled dataset is essential. However, obtaining such a dataset is particularly challenging in monocular threedimensional (3D) tracking of transient objects, like welding spatters, which are so complex that no single model can accurately describe them simultaneously. To address the label acquisition issue in this transient process, we introduce a models-driven weakly supervised learning (MsWSL) approach which is achieved by integrating diverse scopes of multiple models and generating multiple labels for weakly supervised learning. The MsWSL is incorporated into a combination of a feature image transformer encoder and a multilayer perceptron with multiple inputs, enhancing the network's ability to automatically adjust the model and adaptively learn the most suitable label. In the experiments on welding spatters, results from MsWSL show better performance than those from individual models, validating the effectiveness and superiority of MsWSL. We believe that the proposed MsWSL will have broader significant applications in monocular 3D-trajectory reconstruction for multiple, intricate, unrepeatable, and transient objects.
Keyword:
Monocular 3D vision
Models-driven
Weakly supervised learning
Welding spatters
3D-trajectory reconstruction
AI总结
对已上传原文的论文进行重点信息的提取，主要内容包括：简要概述、研究摘要、背景介绍、关键亮点、图文解析、展望与总结。
期刊
IF:
3.7
论文数:
7.3K
被引数:
1.7W
机构
引用论文
.png){kind=icon}
分享收藏Generation of hydroxyl radicals by urban suspended particulate air matter. The role of iron ions城市悬浮颗粒物产生羟基自由基。铁离子的作用:
PREAI
分享收藏分享收藏Weakly Supervised Deep Nuclei Segmentation Using Partial Points Annotation in Histopathology Images在组织病理学图像中使用部分点注释进行弱监督的深核分割OAAI
OAAI分享收藏 没有更多内容
没有更多内容