convolution in tensorflow (tf.nn.conv)
来源:互联网 发布:黑色紧身小脚裤淘宝 编辑:程序博客网 时间:2024/05/22 09:02
Understanding convolution in tensorflow
Points should be noticed
The convolution ops sweep a 2-D filter over a batch of images, applying the filter to each window of each image of the appropriate size. The different ops trade off between generic vs. specific filters:
conv2d: Arbitrary filters that can mix channels together.
depthwise_conv2d: Filters that operate on each channel independently.
separable_conv2d: A depthwise spatial filter followed by a pointwise
filter.
from:
Neural Network | TensorFlow
https://www.tensorflow.org/api_guides/python/nn#Convolution
Explanation of the equation in API of tf.nn.conv2d
b
in yellow shadow is for travering through the batch
with images.
the variables in green shadow is for locating a patch
.
In the equation, di
and dj
loop variable to traverse through the height and width of the patch
in one image.
~~I’m not very sure about this. q
is for traversing through the channels of the image, with the step of strike[3]. ~~
ref:
tf.nn.conv2d | TensorFlow
https://www.tensorflow.org/api_docs/python/tf/nn/conv2d
tf.nn.conv[1d, 2d, 3d]
There are tf.nn.conv1d
, tf.nn.conv2d
and tf.nn.conv3d
. Problems are there, which to be choosed and what’s the difference.
I have not studied this deep. For I am using tensorflow to processing images, it seems tf.nn.conv2d
is for me. From my perspective, the difference of them are as following.
tf.nn.conv1d
is for some linear structure, text, sound or some others.tf.nn.conv2d
is for structure appeared to be 2-D, such as image.tf.nn.conv3d
is for something not familiar to me. It seems to be applied in signal processing.
- convolution in tensorflow (tf.nn.conv)
- TensorFlow tf.nn.conv2d
- TensorFlow-tf.nn.conv2d 函数
- Tensorflow系列:tf.nn.conv2d
- 【Tensorflow】tf.nn.relu函数
- 【Tensorflow】tf.nn.dropout函数
- tensorflow函数 tf.nn.xw_plus_b
- Tensorflow学习---tf.nn.embedding_lookup
- TensorFlow函数:tf.nn.in_top_k()
- TensorFlow中 tf.nn.embedding_lookup
- TensorFlow tf.nn.conv2d()介绍
- Tensorflow tf.nn模块分析
- tensorflow源码 tf.nn.softmax_cross_entropy_with_logits & tf.nn.sparse_softmax_cross_entropy_with_log
- tensorflow-BatchNormalization(tf.nn.moments及tf.nn.batch_normalization)
- Tensorflow -- tf.nn.conv2d() 函数详解
- 【TensorFlow】tf.nn.softmax_cross_entropy_with_logits的用法
- tensorflow tf.nn.embedding_lookup(embeddings, train_inputs)解释
- tensorflow中的tf.nn这类函数
- Unity3D-VR手柄控制器的实现
- Django学习(1)开发环境的搭建
- 【Shiro权限管理】6.Shiro认证思路分析
- JavaScript--《剑指offer》-题四
- 从贝叶斯方法谈到贝叶斯网络
- convolution in tensorflow (tf.nn.conv)
- winform程序中使用js
- maven中jetty插件配置,注意不同jetty对应不同jdk版本
- E
- OpenCV学习记录一:图像的读取与显示
- POJ-1050 To the Max 二维最大子段和
- Android设计模式之——备忘录模式
- 计算系数
- java获得文件路径