Skip to content

Commit 71d5b35

Browse files
authored
docs: update mnn blog
1 parent e4d0e36 commit 71d5b35

File tree

1 file changed

+7
-19
lines changed

1 file changed

+7
-19
lines changed

docs/blog/posts/about_model/support_mnn_engine.md

Lines changed: 7 additions & 19 deletions
Original file line numberDiff line numberDiff line change
@@ -28,22 +28,21 @@ hide:
2828
rapidocr==3.5.0
2929
mnn==3.2.0
3030
text_det_metric==0.0.8
31+
text_rec_metric==0.0.1
3132
datasets==3.6.0
3233
```
3334

34-
## 支持Det模型
35+
## 安装MNN
3536

3637
!!! note
3738

3839
MNN部分主要参考官方文档:https://mnn-docs.readthedocs.io/en/latest/tools/convert.html
3940

40-
### 安装MNN
41-
4241
```bash linenums="1"
4342
pip install MNN
4443
```
4544

46-
### 转换ONNX模型
45+
## 转换ONNX模型
4746

4847
转换命令:
4948

@@ -65,22 +64,9 @@ outputTensors: [ fetch_name_0, ]
6564
Converted Success!
6665
```
6766

68-
### 比较转化前后推理精度差异
69-
70-
#### 运行环境
71-
72-
- 操作系统:macOS Tahoe 26.2
73-
- 硬件:MacBookPro M2
74-
- Python环境:
75-
76-
```bash linenums="1"
77-
rapidocr==3.5.0
78-
mnn==3.2.0
79-
text_det_metric==0.0.8
80-
datasets==3.6.0
81-
```
67+
## 支持Det模型
8268

83-
#### 运行推理代码
69+
### 比较转化前后推理精度差异
8470

8571
两者对比的模型为`ch_PP-OCRv4_det_infer`模型,Exp1基于`RapidOCR+ONNXRuntime`格式模型推理, Exp2基于`RapidOCR+MNN`格式模型推理。
8672

@@ -212,3 +198,5 @@ Converted Success!
212198
|||||||||
213199
|11|ch_PP-OCRv5_server_det|RapidOCR|ONNXRuntime|0.7394|0.8442|0.7883|2.0193|
214200
|12|ch_PP-OCRv5_server_det|RapidOCR|MNN|0.7394|0.8442|0.7883|1.6048|
201+
202+
## 支持Rec模型

0 commit comments

Comments
 (0)