An End-to-End Trainable Neural Network for Image-based Sequence Recognition and Its Application to Scene Text Recognition

Results of accuracy evaluation with tools/eval at different text recognition datasets.

*: 'FP16' or 'INT8' stands for 'model quantized into FP16' or 'model quantized into int8'

Note:

• Model source:
  • text_recognition_CRNN_EN_2021sep.onnx: https://docs.opencv.org/4.5.2/d9/d1e/tutorial_dnn_OCR.html (CRNN_VGG_BiLSTM_CTC.onnx)
  • text_recognition_CRNN_CH_2021sep.onnx: https://docs.opencv.org/4.x/d4/d43/tutorial_dnn_text_spotting.html (crnn_cs.onnx)
  • text_recognition_CRNN_CN_2021nov.onnx: https://docs.opencv.org/4.5.2/d4/d43/tutorial_dnn_text_spotting.html (crnn_cs_CN.onnx)
• text_recognition_CRNN_EN_2021sep.onnx can detect digits (0~9) and letters (return lowercase letters a~z) (see CHARSET_EN_36 for details in crnn.py).
• text_recognition_CRNN_CH_2021sep.onnx can detect digits (0~9), upper/lower-case letters (a~z and A~Z), and some special characters (see CHARSET_CH_94 for details in crnn.py).
• text_recognition_CRNN_CN_2021nov.onnx can detect digits (0~9), upper/lower-case letters (a~z and A~Z), some Chinese characters and some special characters (see CHARSET_CN_3944 for details in crnn.py).
• For details on training this model series, please visit https://github.com/zihaomu/deep-text-recognition-benchmark.
• text_recognition_CRNN_XX_2021xxx_int8bq.onnx represents the block-quantized version in int8 precision and is generated using block_quantize.py with block_size=64.

NOTE:

• This demo uses text_detection_db as text detector.

Run the demo detecting English:

# detect on camera input
python demo.py
# detect on an image
python demo.py --input /path/to/image -v

# get help regarding various parameters
python demo.py --help

Run the demo detecting Chinese:

# detect on camera input
python demo.py --model text_recognition_CRNN_CN_2021nov.onnx
# detect on an image
python demo.py --input /path/to/image --model text_recognition_CRNN_CN_2021nov.onnx

# get help regarding various parameters
python demo.py --help

Install latest OpenCV and CMake >= 3.24.0 to get started with:

# detect on camera input
./build/opencv_zoo_text_recognition_crnn 
# detect on an image
./build/opencv_zoo_text_recognition_crnn --input /path/to/image -v

# get help regarding various parameters
./build/opencv_zoo_text_recognition_crnn --help

Run the demo detecting Chinese:

# detect on camera input
./build/opencv_zoo_text_recognition_crnn  --model=text_recognition_CRNN_CN_2021nov.onnx --charset=charset_3944_CN.txt
# detect on an image
./build/opencv_zoo_text_recognition_crnn  --input=/path/to/image --model=text_recognition_CRNN_CN_2021nov.onnx --charset=charset_3944_CN.txt

# get help regarding various parameters
./build/opencv_zoo_text_recognition_crnn  --help

All files in this directory are licensed under Apache 2.0 License.

• https://arxiv.org/abs/1507.05717
• https://github.com/bgshih/crnn
• https://github.com/meijieru/crnn.pytorch
• https://github.com/zihaomu/deep-text-recognition-benchmark
• https://docs.opencv.org/4.5.2/d9/d1e/tutorial_dnn_OCR.html