提交 17e5df1d 编写于 作者: L LDOUBLEV

mv en.doc to ./doc_en

上级 fe35728b
# BENCHMARK
This document gives the performance of the series models for Chinese and English recognition.
## TEST DATA
We collected 300 images for different real application scenarios to evaluate the overall OCR system, including contract samples, license plates, nameplates, train tickets, test sheets, forms, certificates, street view images, business cards, digital meter, etc. The following figure shows some images of the test set.
<div align="center">
<img src="../datasets/doc.jpg" width = "800" height = "200" />
</div>
## MEASUREMENT
Explanation:
- v1.0 indicates DB+CRNN models without the strategies. v1.1 indicates the PP-OCR models with the strategies and the direction classify. slim_v1.1 indicates the PP-OCR models with prunner or quantization.
- The long size of the input for the text detector is 960.
- The evaluation time-consuming stage is the complete stage from image input to result output, including image pre-processing and post-processing.
- ```Intel Xeon 6148``` is the server-side CPU model. Intel MKL-DNN is used in the test to accelerate the CPU prediction speed.
- ```Snapdragon 855``` is a mobile processing platform model.
Compares the model size and F-score:
| Model Name | Model Size <br> of the <br> Whole System\(M\) | Model Size <br>of the Text <br> Detector\(M\) | Model Size <br> of the Direction <br> Classifier\(M\) | Model Size<br>of the Text <br> Recognizer \(M\) | F\-score |
|:-:|:-:|:-:|:-:|:-:|:-:|
| ch\_ppocr\_mobile\_v1\.1 | 8\.1 | 2\.6 | 0\.9 | 4\.6 | 0\.5193 |
| ch\_ppocr\_server\_v1\.1 | 155\.1 | 47\.2 | 0\.9 | 107 | 0\.5414 |
| ch\_ppocr\_mobile\_v1\.0 | 8\.6 | 4\.1 | \- | 4\.5 | 0\.393 |
| ch\_ppocr\_server\_v1\.0 | 203\.8 | 98\.5 | \- | 105\.3 | 0\.4436 |
Compares the time-consuming on T4 GPU (ms):
| Model Name | Overall | Text Detector | Direction Classifier | Text Recognizer |
|:-:|:-:|:-:|:-:|:-:|
| ch\_ppocr\_mobile\_v1\.1 | 137 | 35 | 24 | 78 |
| ch\_ppocr\_server\_v1\.1 | 204 | 39 | 25 | 140 |
| ch\_ppocr\_mobile\_v1\.0 | 117 | 41 | \- | 76 |
| ch\_ppocr\_server\_v1\.0 | 199 | 52 | \- | 147 |
Compares the time-consuming on CPU (ms):
| Model Name | Overall | Text Detector | Direction Classifier | Text Recognizer |
|:-:|:-:|:-:|:-:|:-:|
| ch\_ppocr\_mobile\_v1\.1 | 421 | 164 | 51 | 206 |
| ch\_ppocr\_mobile\_v1\.0 | 398 | 219 | \- | 179 |
Compares the model size, F-score, the time-consuming on SD 855 of between the slim models and the original models:
| Model Name | Model Size <br> of the <br> Whole System\(M\) | Model Size <br>of the Text <br> Detector\(M\) | Model Size <br> of the Direction <br> Classifier\(M\) | Model Size<br>of the Text <br> Recognizer \(M\) | F\-score | SD 855<br>\(ms\) |
|:-:|:-:|:-:|:-:|:-:|:-:|:-:|
| ch\_ppocr\_mobile\_v1\.1 | 8\.1 | 2\.6 | 0\.9 | 4\.6 | 0\.5193 | 306 |
| ch\_ppocr\_mobile\_slim\_v1\.1 | 3\.5 | 1\.4 | 0\.5 | 1\.6 | 0\.521 | 268 |
# BENCHMARK
This document gives the prediction time-consuming benchmark of PaddleOCR Ultra Lightweight Chinese Model (8.6M) on each platform.
This document gives the performance of the series models for Chinese and English recognition.
## TEST DATA
* 500 images were randomly sampled from the Chinese public data set [ICDAR2017-RCTW](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/doc/doc_ch/datasets.md#ICDAR2017-RCTW-17).
Most of the pictures in the set were collected in the wild through mobile phone cameras.
Some are screenshots.
These pictures show various scenes, including street scenes, posters, menus, indoor scenes and screenshots of mobile applications.
## MEASUREMENT
The predicted time-consuming indicators on the four platforms are as follows:
We collected 300 images for different real application scenarios to evaluate the overall OCR system, including contract samples, license plates, nameplates, train tickets, test sheets, forms, certificates, street view images, business cards, digital meter, etc. The following figure shows some images of the test set.
<div align="center">
<img src="../datasets/doc.jpg" width = "800" height = "200" />
</div>
| Long size(px) | T4(s) | V100(s) | Intel Xeon 6148(s) | Snapdragon 855(s) |
| :---------: | :-----: | :-------: | :------------------: | :-----------------: |
| 960 | 0.092 | 0.057 | 0.319 | 0.354 |
| 640 | 0.067 | 0.045 | 0.198 | 0.236 |
| 480 | 0.057 | 0.043 | 0.151 | 0.175 |
## MEASUREMENT
Explanation:
* The evaluation time-consuming stage is the complete stage from image input to result output, including image
pre-processing and post-processing.
* ```Intel Xeon 6148``` is the server-side CPU model. Intel MKL-DNN is used in the test to accelerate the CPU prediction speed.
To use this operation, you need to:
* Update to the latest version of PaddlePaddle: https://www.paddlepaddle.org.cn/documentation/docs/zh/install/Tables.html#whl-dev
Please select the corresponding mkl version wheel package according to the CUDA version and Python version of your environment,
for example, CUDA10, Python3.7 environment, you should:
```
# Obtain the installation package
wget https://paddle-wheel.bj.bcebos.com/0.0.0-gpu-cuda10-cudnn7-mkl/paddlepaddle_gpu-0.0.0-cp37-cp37m-linux_x86_64.whl
# Installation
pip3.7 install paddlepaddle_gpu-0.0.0-cp37-cp37m-linux_x86_64.whl
```
* Use parameters ```--enable_mkldnn True``` to turn on the acceleration switch when making predictions
* ```Snapdragon 855``` is a mobile processing platform model.
- v1.0 indicates DB+CRNN models without the strategies. v1.1 indicates the PP-OCR models with the strategies and the direction classify. slim_v1.1 indicates the PP-OCR models with prunner or quantization.
- The long size of the input for the text detector is 960.
- The evaluation time-consuming stage is the complete stage from image input to result output, including image pre-processing and post-processing.
- ```Intel Xeon 6148``` is the server-side CPU model. Intel MKL-DNN is used in the test to accelerate the CPU prediction speed.
- ```Snapdragon 855``` is a mobile processing platform model.
Compares the model size and F-score:
| Model Name | Model Size <br> of the <br> Whole System\(M\) | Model Size <br>of the Text <br> Detector\(M\) | Model Size <br> of the Direction <br> Classifier\(M\) | Model Size<br>of the Text <br> Recognizer \(M\) | F\-score |
|:-:|:-:|:-:|:-:|:-:|:-:|
| ch\_ppocr\_mobile\_v1\.1 | 8\.1 | 2\.6 | 0\.9 | 4\.6 | 0\.5193 |
| ch\_ppocr\_server\_v1\.1 | 155\.1 | 47\.2 | 0\.9 | 107 | 0\.5414 |
| ch\_ppocr\_mobile\_v1\.0 | 8\.6 | 4\.1 | \- | 4\.5 | 0\.393 |
| ch\_ppocr\_server\_v1\.0 | 203\.8 | 98\.5 | \- | 105\.3 | 0\.4436 |
Compares the time-consuming on T4 GPU (ms):
| Model Name | Overall | Text Detector | Direction Classifier | Text Recognizer |
|:-:|:-:|:-:|:-:|:-:|
| ch\_ppocr\_mobile\_v1\.1 | 137 | 35 | 24 | 78 |
| ch\_ppocr\_server\_v1\.1 | 204 | 39 | 25 | 140 |
| ch\_ppocr\_mobile\_v1\.0 | 117 | 41 | \- | 76 |
| ch\_ppocr\_server\_v1\.0 | 199 | 52 | \- | 147 |
Compares the time-consuming on CPU (ms):
| Model Name | Overall | Text Detector | Direction Classifier | Text Recognizer |
|:-:|:-:|:-:|:-:|:-:|
| ch\_ppocr\_mobile\_v1\.1 | 421 | 164 | 51 | 206 |
| ch\_ppocr\_mobile\_v1\.0 | 398 | 219 | \- | 179 |
Compares the model size, F-score, the time-consuming on SD 855 of between the slim models and the original models:
| Model Name | Model Size <br> of the <br> Whole System\(M\) | Model Size <br>of the Text <br> Detector\(M\) | Model Size <br> of the Direction <br> Classifier\(M\) | Model Size<br>of the Text <br> Recognizer \(M\) | F\-score | SD 855<br>\(ms\) |
|:-:|:-:|:-:|:-:|:-:|:-:|:-:|
| ch\_ppocr\_mobile\_v1\.1 | 8\.1 | 2\.6 | 0\.9 | 4\.6 | 0\.5193 | 306 |
| ch\_ppocr\_mobile\_slim\_v1\.1 | 3\.5 | 1\.4 | 0\.5 | 1\.6 | 0\.521 | 268 |
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册