ID 原文 译文
52547 该方法使用元路径对异构图中涉及到的语义关系进行建模，并利用图卷积模块和语义级别的注意力机制来捕获单个节点的局部表征。 The model uses the meta-path to model the structure involving semantics in heterogeneous graphs and utilizes graph convolution module and semantic-level attention mechanism to capture individual node local representations.
52548 该方法通过最大化单个节点与局部子图间的互信息，有效地学习高阶节点表征。 By maximizing the mutual information between the individual node embedding and the local graph, the proposed model effectively learns high-level node representations.
52549 实验结果表明，该方法相比基于全局图互信息的方法，可以将数据集DBLP/IMDB上的节点分类任务的微值F1(micro-F1)提高大约3%/9%，同时将DBLP/IMDB上的节点聚类任务的调整兰德系数(ARI)提高约23%/46%。 The experimental results show that compared with HDGI which is based on the global graph mutual information maximization, the proposed method can increase the micro-F1 of the node classification task on DBLP/IMDB up to about 3%/9%, as well as the adjusted Rand index(ARI) of the node clustering task on DBLP/IMDB up to about 23%/46%.
52550 密封圈的缺陷区域过小、缺陷特征与背景相似度较高和密封圈上杂质过多，导致密封圈表面缺陷检测难度很大。 Due to the undersized defect areas of the sealing ring, the high similarity between the defect feature and the background and the redundancy of impurities on the sealing ring, the detection of the sealing ring's surface defects has been extremely difficult.
52551 对此，本文提出一种基于编解码结构的语义分割缺陷检测算法。 Aiming at solving the problems faced by defect detection, this paper proposes a semantic segmentation defect detection algorithm based on encoding/decoding structure.
52552 首先，通过向编码器端的各个卷积模块注入不同尺度的图像特征，以此增加模型的多尺度信息和冗余信息， Firstly, image features at different scales are inserted into each convolution module on the encoder side to increase the multi-scale information and redundant information of the model.
52553 同时使用膨胀卷积层替换常规卷积层，有效地增加网络的感受野，使得模型不易漏掉小的缺陷; At the same time, the dilated convolution layer is used to replace the conventional convolution layer, therefore effectively increasing the receptive field of the network and decreasing the risk of missing small defects.
52554 其次，结合注意力机制将编码器的信息与解码器端的信息进行级联，加强模型对目标区域信息的捕获; Secondly, combined with the attention mechanism, the information of the encoder and the decoder is cascaded, and the capture of the target area's information by this model is optimized.
52555 最后，将解码器端输出的不同尺度信息进行融合，提高网络对小缺陷的分割精度。 Finally, the fusion of the different-scaled information from the decoder could improve the network' s segmentation accuracy for small defects.
52556 实验结果表明，本文所提出的方法能够对较小的、边缘不清晰的密封圈缺陷进行有效分割，能够满足密封圈缺陷的检测需求。 The experimental results show that the proposed method can effectively divide the small and unclear seal ring defects, and can meet the requirements of the detection of seal ring defects.