search

sergiomsilva / alpr-unconstrained

License Plate Detection and Recognition in Unconstrained Scenarios
Other
1.71k stars 607 forks source link

Could release the OCR training code? #61

Open xiaowenhe opened 5 years ago

xiaowenhe commented 5 years ago

hello ,@sergiomsilva, could you release the OCR training code now? I follow your WPOD-NET, and get good results in our datasets, but this is only LP detection. And OCR result is not well. So hope you can release the OCR training code. Thank you very much!

NNDam commented 5 years ago

hello ,@sergiomsilva, could you release the OCR training code now? I follow your WPOD-NET, and get good results in our datasets, but this is only LP detection. And OCR result is not well. So hope you can release the OCR training code. Thank you very much!

For OCR training, follow the darknet training tutorial:

TyroneLi commented 5 years ago

hello ,@sergiomsilva, could you release the OCR training code now? I follow your WPOD-NET, and get good results in our datasets, but this is only LP detection. And OCR result is not well. So hope you can release the OCR training code. Thank you very much!

For OCR training, follow the darknet training tutorial:

  • Add " train " and " valid " lines to ocr-net.data
  • Train command example: ./darknet detector train data/ocr/ocr-net.data data/ocr/ocr-net.cfg ocr-net.weights

Hi~ @NNDam ,for ocr-net trainng and follow the steps according to offical yolo framwork,do I need to create specific format dataset,which means I have to label each letter or digit inside a plate manually?Or just give one label txt file including all letters and digits inside that plate?Which method should I use to create custom dataset so I can re-train ocr-net using yolo?Thanls~

NNDam commented 5 years ago

@TyroneLi You have to label each letter and digits inside plate manually, like labelImg-tool

TyroneLi commented 5 years ago

@TyroneLi You have to label each letter and digits inside plate manually, like labelImg-tool

@NNDam Okay,I know how to do.So that is like the normal steps to train the official yolo algorithm to detect each object in the image.Before that I was really confused about the ocr-net--how can it segment and detect each letter and digit in the plate at once.Now I know they just use detection algorithm idea.Thanks~