随笔 - 884,  文章 - 0,  评论 - 38,  阅读 - 155万

随笔分类 -  Tensorflow

yolo——tensorflow实现
该文被密码保护。
posted @ 2018-12-31 10:17 一抹烟霞 阅读(2) 评论(0) 推荐(0) 编辑
(Ubuntu)Tensorflow object detection API——(3)创建训练/测试数据集
摘要:1、下载labelImg工具进行标注 https://github.com/tzutalin/labelImg (1)点击打开训练图片所在的文件夹 (2)点击框选自己要识别的目标 (3)添加标签并保存,获得同名的xml文件,如图。 2、将文件夹内的xml文件内的信息统一记录到.csv表格中 # Au 阅读全文
posted @ 2018-12-25 21:32 一抹烟霞 阅读(150) 评论(0) 推荐(0) 编辑
(Ubuntu)Tensorflow object detection API——(2)运行已经训练好的模型
摘要:1、在models/research/文件下运行命令: jupyter-notebook 在弹出的页面里点击 http://localhost:8888/?token=6d343b91cce1d30e8b3d13fbaa3fd9d0e86a9cd770c8ce26 2、打开object_detect 阅读全文
posted @ 2018-12-24 23:52 一抹烟霞 阅读(169) 评论(0) 推荐(0) 编辑
tensorflow object detection API 验证时报No module named 'object_detection'
摘要:1、Windows下: (1)在 你的Anaconda3安装路/Anaconda3/Lib/site-packages 下新建一个txt文件 (我这里的安装路径是C:\ProgramData\Anaconda3\Lib\site-packages) (2)在新建的txt文件中写入电脑上\models 阅读全文
posted @ 2018-12-24 23:39 一抹烟霞 阅读(388) 评论(0) 推荐(0) 编辑
(Ubuntu)Tensorflow object detection API——(1)环境搭建
摘要:1、下载Tensorflow object detection API : git clone https://github.com/tensorflow/models.git 2、安装依赖 Protobuf >= 3.0.0 Python-tk Pillow lxml tf Slim Jupyte 阅读全文
posted @ 2018-12-24 22:37 一抹烟霞 阅读(156) 评论(0) 推荐(0) 编辑
《21个项目》书中第三章data_convert.py文件运行失败解决办法
摘要:问题原因:代码是用python2写的,当用python3时会报错。 把代码改成如下python3的格式: data_convert.py # coding:utf-8 from __future__ import absolute_import import argparse import os i 阅读全文
posted @ 2018-12-22 22:49 一抹烟霞 阅读(198) 评论(0) 推荐(0) 编辑
将图片数据保存为单个tfrecord文件
摘要:# Author Qian Chenglong # -*-coding: utf-8 -*- import tensorflow as tf import numpy as np import os import cv2 import matplotlib.pyplot as plt import 阅读全文
posted @ 2018-12-22 22:35 一抹烟霞 阅读(203) 评论(0) 推荐(0) 编辑
Tensorflow把label转化成one_hot格式
摘要:one_hot( indices, #指数的张量,即label的矩阵形式(1行 n列) depth, #一个标量,用于定义一个 one hot 维度的深度。即分类数 on_value=None, #定义在 indices[j] = i 时填充输出的值的标量。(默认:1) off_value=None 阅读全文
posted @ 2018-12-19 00:15 一抹烟霞 阅读(731) 评论(0) 推荐(0) 编辑
Tensorflow——tf.nn.max_pool池化操作
摘要:tf.nn.max_pool( value#需要池化的输入,一般池化层接在卷积层后面,所以输入通常是feature map,依然是[batch, height, width, channels]这样的shape ksize,#池化窗口的大小,取一个四维向量,一般是[1, height, width, 阅读全文
posted @ 2018-12-17 23:41 一抹烟霞 阅读(202) 评论(0) 推荐(0) 编辑
Tensorflow——tf.nn.conv2d卷积操作
摘要:tf.nn.conv2d( input, #指需要做卷积的输入图像,给定形状为[batch, in_height, in_width, in_channels]的输入张量。具体含义是[训练时一个batch的图片数量, 图片高度, 图片宽度, 图像通道数] filter, #形状为[filter_he 阅读全文
posted @ 2018-12-17 23:13 一抹烟霞 阅读(175) 评论(0) 推荐(0) 编辑
Tensoflow简单神经网络实现非线性拟合
摘要:定义一个自动增加网络层数的函数 权重weight的设置:在生成初始参数时,随机变量(normal distribution)会比全部为0要好很多,所以我们这里的weights为一个in_size行, out_size列的随机变量矩阵。 Weights = tf.Variable(tf.random_ 阅读全文
posted @ 2018-12-17 21:00 一抹烟霞 阅读(318) 评论(0) 推荐(0) 编辑
Tensorflow 中的 placeholder
摘要:placeholder 是 Tensorflow 中的占位符,暂时储存变量. Tensorflow 中用于从外部传入data的占位符, 用input= tf.placeholder()设置占位符, 然后以这种形式传输数据 sess.run(***, feed_dict={input: **}). 注 阅读全文
posted @ 2018-12-16 00:53 一抹烟霞 阅读(636) 评论(0) 推荐(0) 编辑
Tensorflow 中的constant、Session、placeholde和 Variable简介
摘要:Tensorflow 中的constant和 Variabletf.constant()Sessiontf.placeholder()Session’s feed_dicttf.Variable()初始化 在 TensorFlow 中,数据不是以整数,浮点数或者字符串形式存在的。这些值被封装在一个叫 阅读全文
posted @ 2018-12-16 00:46 一抹烟霞 阅读(333) 评论(0) 推荐(0) 编辑
TensorFlow Training函数
该文被密码保护。
posted @ 2018-12-15 22:11 一抹烟霞 阅读(0) 评论(0) 推荐(0) 编辑
Tensorflow图像数据读入
该文被密码保护。
posted @ 2018-12-13 22:21 一抹烟霞 阅读(0) 评论(0) 推荐(0) 编辑
Tensorflow——API说明
该文被密码保护。
posted @ 2018-12-07 11:06 一抹烟霞 阅读(0) 评论(0) 推荐(0) 编辑
Tensorflow——实现自编码器
该文被密码保护。
posted @ 2018-12-04 15:43 一抹烟霞 阅读(3) 评论(0) 推荐(0) 编辑
Windows下安装Tensorflow—GPU版本
摘要:https://blog.csdn.net/weixin_39290638/article/details/80045236 阅读全文
posted @ 2018-12-03 20:38 一抹烟霞 阅读(84) 评论(0) 推荐(0) 编辑
生成验证码
该文被密码保护。
posted @ 2018-12-01 23:59 一抹烟霞 阅读(0) 评论(0) 推荐(0) 编辑
Tensorflow——使用inception-v3做各种图像的识别
该文被密码保护。
posted @ 2018-12-01 12:23 一抹烟霞 阅读(1) 评论(0) 推荐(0) 编辑

< 2025年3月 >
23 24 25 26 27 28 1
2 3 4 5 6 7 8
9 10 11 12 13 14 15
16 17 18 19 20 21 22
23 24 25 26 27 28 29
30 31 1 2 3 4 5

Live2D
欢迎阅读『Tensorflow』

喜欢请打赏

扫描二维码打赏

了解更多

点击右上角即可分享
微信分享提示