ima

html2canvas and toDataURL generated image has horizontal line

匿名 (未验证) 提交于 2019-12-03 01:00:01
可以将文章内容翻译成中文,广告屏蔽插件可能会导致该功能失效(如失效,请关闭广告屏蔽插件后再试): 问题: I am looping through 10-14 html elements, and generating image data in an array for later use for insertion into a PDF. The problem is that these images occasionally have a horizontal line across them, which seems seems to be an existing issue with html2canvas. Happens mostly in FF and IE, and occationally on Chrome, but not as often. function convertElementsToImages(containerSelector) { if (_this.pdfDebug) { window.console.log('convertElementsToImages'); window.console.log(containerSelector); } var eles = $(containerSelector + ' > *'), q =

论文:Deep Residual Learning for Image Recognition

匿名 (未验证) 提交于 2019-12-03 00:20:01
Abstract: 更深层的网络训练十分困难,我们提出了残差网络来实现深层网络。我们重新定制了层间的学习是参考 layer input 的残差函数,而不是一个没有参考的函数。 Introduction: 是否学习更好的网络就是简单的堆积更多的层?一个障碍便是梯度消失或者爆炸,从训练的一开始便会损害收敛,虽然这个问题可以被 normalization initialization 和 intermediate normalization layers解决,可以使网络达到十多层 当深层网络可以收敛后,又出现了退化问题,当网络层数加深,准确率开始变得饱和,然后会快速衰退,这种衰退并不是因为过拟合引起的,加更多的层会引起训练误差变大,如下图 这种训练准确率的退化表明并不是所用的 system 都可以很好的优化。如何从 shallower architecture 到 deeper architecture. 现有的方法是,added layers 是恒等映射(identity mapping),而其它层是 copy from learned shallower model.这种方法本应获得不低于 shallower counterpart 的准确率,但是并没有。 本文通过提出一个 deep residual learning framework 来解决 degradation

安装部署k8s-版本-1.13

匿名 (未验证) 提交于 2019-12-02 23:43:01
# 统一hosts cat /etc/hosts 10.0.0.10 k8s-master 10.0.0.20 k8s-node1 10.0.0.30 k8s-node2 # 同步时间 ntpdate ntp1.aliyun.com # 禁用SELINUX,关闭firewalld和iptables.service,k8s会自己设置iptables网络策略 systemctl stop iptables.service systemctl disable iptables.service systemctl stop firewalld.service systemctl disable firewalld.service # 关闭swap,修改/etc/fstab文件,注释掉SWAP的自动挂载,使用free -m确认swap已经关闭 swapoff -a # 调整swappiness参数 # 创建/etc/sysctl.d/k8s.conf文件,添加如下内容 net.bridge.bridge-nf-call-ip6tables = 1 net.bridge.bridge-nf-call-iptables = 1 net.ipv4.ip_forward = 1 vm.swappiness=0 # 生效 modprobe br_netfilter sysctl -p /etc

OpenCV && C++ 09 - Gaussian Blur on Images with OpenCV

六眼飞鱼酱① 提交于 2019-11-30 07:04:16
3 x 3 Gaussian Kernel 5 x 5 Gaussian Kernel Code /* 作者:郑大峰 时间:2019年09月23日 环境:OpenCV 4.1.1 + VS2017 内容:Gaussian Blur on Images with OpenCV */ #include "pch.h" #include <iostream> #include <opencv2/opencv.hpp> using namespace std; using namespace cv; int main() { Mat image = imread("claudia.png"); if (image.empty()) { cout << "Could not open or find the image" << endl; cin.get(); return -1; } //Blur the image with 3x3 Gaussian kernel Mat image_blurred_with_3x3_kernel; GaussianBlur(image, image_blurred_with_3x3_kernel, Size(3, 3), 0); //Blur the image with 5x5 Gaussian kernel Mat image_blurred