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
afeixing77
Paddle
提交
56936b9e
P
Paddle
项目概览
afeixing77
/
Paddle
与 Fork 源项目一致
Fork自
PaddlePaddle / Paddle
通知
1
Star
0
Fork
0
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
0
列表
看板
标记
里程碑
合并请求
0
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
Paddle
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
0
Issue
0
列表
看板
标记
里程碑
合并请求
0
合并请求
0
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
体验新版 GitCode,发现更多精彩内容 >>
提交
56936b9e
编写于
10月 20, 2018
作者:
D
Dang Qingqing
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Refine doc for generate_proposals_op.
test=develop
上级
4801ee8f
变更
1
隐藏空白更改
内联
并排
Showing
1 changed file
with
36 addition
and
24 deletion
+36
-24
paddle/fluid/operators/detection/generate_proposals_op.cc
paddle/fluid/operators/detection/generate_proposals_op.cc
+36
-24
未找到文件。
paddle/fluid/operators/detection/generate_proposals_op.cc
浏览文件 @
56936b9e
...
...
@@ -453,33 +453,45 @@ class GenerateProposalsKernel : public framework::OpKernel<T> {
class
GenerateProposalsOpMaker
:
public
framework
::
OpProtoAndCheckerMaker
{
public:
void
Make
()
override
{
AddInput
(
"Scores"
,
"The scores of anchors should be foreground."
);
AddInput
(
"BboxDeltas"
,
"bbox_deltas."
);
AddInput
(
"ImInfo"
,
"Information for image reshape."
);
AddInput
(
"Anchors"
,
"All anchors."
);
AddInput
(
"Variances"
,
" variances"
);
AddOutput
(
"RpnRois"
,
"Anchors."
);
AddOutput
(
"RpnRoiProbs"
,
"Anchors."
);
AddAttr
<
int
>
(
"pre_nms_topN"
,
"pre_nms_topN"
);
AddAttr
<
int
>
(
"post_nms_topN"
,
"post_nms_topN"
);
AddAttr
<
float
>
(
"nms_thresh"
,
"nms_thres"
);
AddAttr
<
float
>
(
"min_size"
,
"min size"
);
AddInput
(
"Scores"
,
"(Tensor) The scores from conv is in shape (N, A, H, W), "
"N is batch size, A is number of anchors, "
"H and W are height and width of the feature map"
);
AddInput
(
"BboxDeltas"
,
"(Tensor) Bounding box deltas from conv is in "
"shape (N, 4*A, H, W)."
);
AddInput
(
"ImInfo"
,
"(Tensor) Information for image reshape is in shape (N, 3), "
"in format (height, width, scale)"
);
AddInput
(
"Anchors"
,
"(Tensor) Bounding box anchors from anchor_generator_op "
"is in shape (A, H, W, 4)."
);
AddInput
(
"Variances"
,
"(Tensor) Bounding box variances with same shape as `Anchors`."
);
AddOutput
(
"RpnRois"
,
"(LoDTensor), Output proposals with shape (rois_num, 4)."
);
AddOutput
(
"RpnRoiProbs"
,
"(LoDTensor) Scores of proposals with shape (rois_num, 1)."
);
AddAttr
<
int
>
(
"pre_nms_topN"
,
"Number of top scoring RPN proposals to keep before "
"applying NMS."
);
AddAttr
<
int
>
(
"post_nms_topN"
,
"Number of top scoring RPN proposals to keep after "
"applying NMS"
);
AddAttr
<
float
>
(
"nms_thresh"
,
"NMS threshold used on RPN proposals."
);
AddAttr
<
float
>
(
"min_size"
,
"Proposal height and width both need to be greater "
"than this min_size."
);
AddAttr
<
float
>
(
"eta"
,
"The parameter for adaptive NMS."
);
AddComment
(
R"DOC(
Generate Proposals OP
This operator Generate bounding box proposals for Faster RCNN.
The propoasls are generated for a list of images based on image
score 'Scores', bounding box regression result 'BboxDeltas' as
well as predefined bounding box shapes 'anchors'. Greedy
non-maximum suppression is applied to generate the final bounding
boxes.
This operator proposes rois according to each box with their probability to be a foreground object and
the box can be calculated by anchors. Bbox_details and scores are the output of RPN. Final proposals
could be used to train detection net.
Scores is the probability for each box to be an object. In format of (N, A, H, W) where N is batch size, A is number
of anchors, H and W are height and width of the feature map.
BboxDeltas is the differece between predicted box location and anchor location. In format of (N, 4*A, H, W)
For generating proposals, this operator transposes and resizes scores and bbox_deltas in size of (H*W*A, 1) and (H*W*A, 4) and
calculate box locations as proposals candidates. Then clip boxes to image and remove predicted boxes with small area.
Finally, apply nms to get final proposals as output.
)DOC"
);
}
};
...
...
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