Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
wangsen
paddle_dbnet
Commits
19250362
Commit
19250362
authored
May 05, 2022
by
LDOUBLEV
Browse files
fix
parent
a5b5bfe4
Changes
1
Hide whitespace changes
Inline
Side-by-side
Showing
1 changed file
with
1 addition
and
1 deletion
+1
-1
doc/doc_ch/PP-OCRv3_introduction.md
doc/doc_ch/PP-OCRv3_introduction.md
+1
-1
No files found.
doc/doc_ch/PP-OCRv3_introduction.md
View file @
19250362
...
...
@@ -46,7 +46,7 @@ RSEFPN的网络结构如下图所示,RSEFPN在PP-OCRv2的FPN基础上,将FPN
RSEFPN将PP-OCR检测模型的精度hmean从81.3%提升到84.5%。模型大小从3M变为3.6M。
*注:PP-OCRv2的FPN通道数仅为96和24,如果直接用SE模块代替FPN的卷积会导致精度下降,
使用
残差结构可以防止训练中包含重要特征的通道被抑制。*
*注:PP-OCRv2的FPN通道数仅为96和24,如果直接用SE模块代替FPN的卷积会导致精度下降,
RSEConv引入
残差结构可以防止训练中包含重要特征的通道被抑制。*
-
在蒸馏的teacher模型精度提升方面,提出了LKPAN结构替换PP-OCRv2的FPN结构,并且使用ResNet50作为Backbone,更大的模型带来更多的精度提升。另外,对teacher模型使用
[
DML
](
https://arxiv.org/abs/1706.00384
)
蒸馏策略进一步提升teacher模型的精度。最终teacher的模型指标相比ppocr_server_v2.0从83.2%提升到了86.0%。
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment