Paddleocr learning (II) paddleocr detection model training

Latest in school PaddleOCR There is little information on the Internet , Most of them are ignorant , Maybe less contact . So I just sort out and write my own materials .

1. install PaddleOCR

github Address

2. function

# Run to root directory 
python PPOCRLabel.py --lang ch

After running, it's like this

The default model will be downloaded from the cloud for the first time
The downloaded model will be displayed in ：

C:\Users\ Your username \.paddleocr

Built in model Default model ：PPOCRLabel By default PaddleOCR Chinese and English super lightweight OCR Model , Support Chinese and English and number recognition , Multilingual detection .

Model language switching ： The user can use the menu bar “PaddleOCR” - “ Choose a model ”
Switch the built-in model language , Currently supported languages include French 、 German 、 Korean 、 Japanese . For specific model download links, please refer to PaddleOCR Model list .

Custom model ： Users can use... According to the custom model code , By modifying the PPOCRLabel.py Targeted at PaddleOCR Class instantiation is replaced by its own training model .

Be careful ：

1.PPOCRLabel It does not support automatic annotation of pictures with Chinese file names 
2.PPOCRLabel Take the folder as the basic marking unit , After opening the picture folder to be marked , The picture will not be displayed in the window bar , Instead, click  “ Select the folder ”  Then directly import the pictures under the folder into the program .
3. The picture status indicates whether the user has manually saved this picture , If it has not been saved manually, it is  “X”, Manually saved as  “√”. Click on  “ Auto Label ” After button ,PPOCRLabel Not for the status of  “√”  Re label your picture .
4. Click on “ Re identification ” after , The model will cover the recognition results in the picture . Therefore, if the recognition result has been manually changed before , There may be changes after re identification .
5.PPOCRLabel The generated file is placed in the marked picture folder , These include the following , Please do not change it manually , Otherwise, the program will be abnormal .

3. Training

First load the training image file , Folder selected here
file - Open Directory - Select a picture folder
Then select auto annotation in the lower left corner
The text in the picture is marked
Right click to delete or mark the wrong place
Then confirm Next And so on