Annie-qu-优快云博客

原创 Django+MySQL+uwsgi+Nginx部署

Django上线服务器配置

2017-08-12 20:07:05 653

原创 Keras载入模型AttributeError: ‘str‘ object has no attribute ‘decode‘

AttributeError: 'str' object has no attribute 'decode'原因：python3里面，字符串要先encode手动指定其为某一编码的字节码之后，才能decode解码。解决：修改源码，找到报错的对应文件，这次是topology.py原本是original_backend = f.attrs['backend'].decode('utf8')改为：original_backend = f.attrs['backend'].encode.

2021-10-26 19:41:24 457

原创 tensorflow-gpu 安装（Windows）

python==3.6tensorflow-gpu==1.13.1 cuda==10.0 cudnn==7.4.1.5keras==2.1.51、官网查看TensorFlow版本对应的cuda版本和cudnn版本，下载TensorFlow与cuda对应：安装cuda：注意显卡是否支持，然后去官网下载对应的cuda，安装。非默认路径（改路径的时候，要改成\NVIDIA GPU Computing Toolkit\CUDA\v10.0这样），加path：比如说像下面这样cud..

2021-10-26 19:03:03 343

转载 Anaconda错误CondaHTTPError: HTTP None None for url

Anaconda换了清华镜像源后还是CondaHTTPError: HTTP None None for url https://mirrors.tuna.tsinghua..._忧郁的康斯卡麦的博客-优快云博客改镜像源 https -->httpC盘下：C:\Users\jiang\.condarcchannels: - defaultsshow_channel_urls: truedefault_channels: - http://mirrors.tuna.tsing

2021-10-26 14:29:43 1530

原创 git使用

Git很久之前就知道要开始版本管理，但是现在才开始。。。首先呢，要下载一个Git bash，安装上。使用见前面命令行1. 配置GitHub(1) 打开 Git Bash(2) 复制并运行 rm -rf ~/.ssh/*把现有的 ssh key 都删掉，这句命令行如果你多打一个空格，可能就要重装系统了，建议复制运行。(3) 运行 ssh-keygen -t rsa -b 4096 -C "你的邮箱"(4) 按回车三次(5) 运行 cat ~/.ssh/id_rsa.pub，里面的内容是SSH

2021-03-21 14:17:39 202

翻译读文献--《U-Net: Convolution Networks for Biomedical Image Segmentation 》

声明：作者翻译论文仅为学习，如有侵权请联系作者删除博文，谢谢！1.AbstractIn this paper, we present a network and training strategy that relies on the strong use of data augmentation to use the available annotated samples more efficiently. The architecture consists of a contracting path

2020-10-30 14:51:17 551

翻译读文献——《Curriculum learning》

原文地址：https://dl.acm.org/doi/10.1145/1553374.15533801、AbstractHumans and animals learn much better when the examples are not randomly presented but organized in a meaningful order which illustrates gradually more concepts, and gradually more complex ones.

2020-08-03 14:56:24 906

翻译读文献——《Very Deep Convolutional Networks for Large-scale Image Recognition》

1. AbstractOur main contribution is a thorough evaluation of increasing depth using an architecture with very small (3×3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth

2020-07-29 15:36:47 442

翻译读文献——《ImageNet classification with deep convolutional neural networks》

1、Introduction当前的对象识别方法中，必不可少的的属机器学习了，而以前万级别的图像数据库已经做得很好了，现在百万级的数据的数据库产生，例如LabelMe, ImageNet。所以开始这数百万级的对象识别研究。However, the immense complexity of the object recognition task means that this problem cannot be specified even by a dataset as large as Image

2020-07-23 11:18:40 657

原创读文献——《Learning representations by back-propagating errors》

Back-procedure, the procedure repeatedly adjusts the weights of the connections in the network so as to minimize a measure of the difference between the actual output vector of the net and the desired output vector. Internal ‘hidden’ units are not part of

2020-07-16 10:40:08 602

原创读文献——《Deep Residual Learning for Image Recognition》

1、AbstractWe present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. (本文提出一种比过去使用的神经网络更深的训练网络，残差学习结构。)We provide residual networks are easier to optimize, and can gain accuracy fr

2020-07-15 10:38:55 362

原创读文献——《Going deeper with convolutions》

Going deeper with convolutions其实这篇文献是在上一个BN那篇文献之前读的，读完这篇日常百度看看相关的时候看到有人推荐四篇相关的文章才去读的BN和ReNet。现在刚看完ResNet，先把这个之前读的回顾一遍再捋一捋ResNet。首先，这是Google在ImageNet2014上的新卷积神经网络GoogLeNet，从名字上可以看出来它和LeNet的密切关系。此外，文中提出了一种新的深度卷积神经网络结构叫做Inception，这是GoogLeNet的关键。文中提到的Heb

2020-07-08 20:56:49 431

翻译读文献——《Batch Normalization Accelerating Deep Network Training by Reducing Internal Covariate Shift》

在自己阅读文章之前，通过网上大神的解读先了解了一下这篇文章的大意，英文不够好的惭愧...大佬的文章在https://blog.youkuaiyun.com/happynear/article/details/44238541，掺杂着我一小些理解和大佬的总结，记录一下我的学习，大家一起交流。首先，引用大佬对文章的梳理解释，大概对本文的理解如下：文章主要理解...

2020-06-29 17:28:13 379

原创慢慢学AI--tensor

Tensor1. 数据类型类型解释torch.IntTensortorch.FloatTensortorch.rand浮点型随机（0,1）区间均匀分布torch.randn浮点，指定维度 N（0，1）分布）torch.range浮点，自定义起止范围（起，止，步长）torch.zeros浮点，维度指定，元素值为02.Tensor运算符号解释torch.abs取绝对值torch.clamp截取上下范围（a,

2020-06-09 16:32:48 300

原创慢慢学AI--初识卷积神经网络

# 卷积神经网络卷积神经网络（Convolutional Neural Networks）简称CNN，标准卷积神经网络架构由卷积层、池化层和全连接层构成。卷积层卷积层主要作用是对输入的数据进行特征提取，卷积核（一个指定窗口大小的扫描器）。卷积前有两种边界像素填充方式：same（全部填充0），valid(不处理）。对输入图像的不进行前期处理和像素填充，可能会导致图像的部分像素点不能被滑动窗口捕捉；same方式可以避免这个问题，让输入图像的全部像素都能被滑动窗口捕捉。卷积操作后：.

2020-06-08 14:31:39 264