Scaling analysis of current influence on Hastelloy surface roughness in electro-polishing process

In this study, a series of Hastelloy tapes were electro-polished, and the dividing method was used to carry out a detailed investigation on the influence of polishing current(I) on root mean square(R_q) at various image scales(L). The electro-polishing is found to be effective mainly at L smaller than 10μm, where the R_q–I relationship could be fitted by an exponential decay function with a residual roughness value. An approximate model of electro-polishing process was established to interpret the exponential decay function. This study provides a quantified insight into the electro-polishing process, which could help to obtain more understanding of its mechanism.

Detection and segmentation of multi-class artifacts in endoscopy

为准确定位内窥镜视频中的人造物,帮助医生提升诊断准确率,引入深度神经网络检测与分割模型,采用特征金字塔与级联R-CNN相互结合的框架,并使用PSPNet结合分类器链的思想,从而解决分割及数据匮乏问题,有效提升性能,并在EAD 2019数据集上取得领先的性能。