Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
PaddlePaddle
PaddleDetection
提交
2d707e32
P
PaddleDetection
项目概览
PaddlePaddle
/
PaddleDetection
大约 1 年 前同步成功
通知
695
Star
11112
Fork
2696
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
184
列表
看板
标记
里程碑
合并请求
40
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
PaddleDetection
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
184
Issue
184
列表
看板
标记
里程碑
合并请求
40
合并请求
40
Pages
分析
分析
仓库分析
DevOps
Wiki
0
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
提交
Issue看板
提交
2d707e32
编写于
8月 30, 2017
作者:
H
hedaoyuan
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Refine the comments.
上级
e967645c
变更
1
隐藏空白更改
内联
并排
Showing
1 changed file
with
18 addition
and
18 deletion
+18
-18
paddle/operators/math/im2col.h
paddle/operators/math/im2col.h
+18
-18
未找到文件。
paddle/operators/math/im2col.h
浏览文件 @
2d707e32
...
...
@@ -29,40 +29,40 @@ enum ColFormat { kCFO = 0, kOCF = 1 };
*
* \param imData Image data.
* \param imShape The shape of imData,
* [input
Channels, inputHeight, inputW
idth].
* [input
_channels, input_height, input_w
idth].
* \param colData Column data.
* \param colShape The shape of colData.
*
* If the template argument Format is kCFO, the shape of colData is:
* [input
Channels, filterHeight, filterWidth, outputHeight, outputW
idth]
* [input
_channels, filter_height, filter_width, output_height, output_w
idth]
* So, it is easy to reshape into a convolution matrix for convolution
* calculation based on matrix multiplication.
* The shape of convolution matrix is [height, width], where the height is equal
* input
Channels * filterHeight * filterW
idth, and the width is equal
* output
Height * outputW
idth.
* input
_channels * filter_height * filter_w
idth, and the width is equal
* output
_height * output_w
idth.
*
* Reshape:
* shape of colData shape of convolution matrix
* [input
C
hannels,
* filter
H
eight,
* filter
W
idth, ======> [height, width]
* output
H
eight,
* output
W
idth]
* [input
_c
hannels,
* filter
_h
eight,
* filter
_w
idth, ======> [height, width]
* output
_h
eight,
* output
_w
idth]
*
* If the template argument Format is kOCF, the shape of colData is:
* [output
Height, outputWidth, inputChannels, filterHeight, filterW
idth]
* [output
_height, output_width, input_channels, filter_height, filter_w
idth]
* So, it is easy to reshape into a sequence matrix for rnn calculation.
* The shape of sequence matrix is [seq
Length, stepSize], where the seqL
ength
* is equal output
Height * outputWidth, and the stepS
ize is equal
* input
Channels * filterHeight * filterW
idth.
* The shape of sequence matrix is [seq
_length, step_size], where the seq_l
ength
* is equal output
_height * output_width, and the step_s
ize is equal
* input
_channels * filter_height * filter_w
idth.
*
* Reshape:
* shape of colData shape of sequence matrix
* [output
H
eight,
* output
W
idth,
* input
C
hannels, ======> [seqLength, stepSize]
* filter
H
eight,
* filter
W
idth]
* [output
_h
eight,
* output
_w
idth,
* input
_c
hannels, ======> [seqLength, stepSize]
* filter
_h
eight,
* filter
_w
idth]
*
* \note The caller needs to ensure that imShape.inputChannels is equal to
* colShape.inputChannels.
...
...
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录