加入CODE CHINA

· 不限速    · 不限空间    · 不限人数    · 私仓免费

免费加入
    README.md

    attention-ocr.pytorch:Encoder+Decoder+attention model

    This repository implements the the encoder and decoder model with attention model for OCR, the encoder uses CNN+Bi-LSTM, the decoder uses GRU. This repository is modified from https://github.com/meijieru/crnn.pytorch
    Earlier I had an open source version, but had some problems identifying images of fixed width. Recently I modified the model to support image recognition with variable width. The function is the same as CRNN. Due to the time problem, there is no pre-training model this time, which will be updated later.

    requirements

    pytorch 0.4.1
    opencv_python

    cd Attention_ocr.pytorch
    pip install -r requirements.txt

    Test

    pretrained model coming soon

    Train

    1. Here i choose a small dataset from Synthetic_Chinese_String_Dataset, about 270000+ images for training, 20000 images for testing. download the image data from Baidu
    2. the train_list.txt and test_list.txt are created as the follow form:
    # path/to/image_name.jpg label
    path/AttentionData/50843500_2726670787.jpg 情笼罩在他们满是沧桑
    path/AttentionData/57724421_3902051606.jpg 心态的松弛决定了比赛
    path/AttentionData/52041437_3766953320.jpg 虾的鲜美自是不可待言
    1. change the trainlist and vallist parameter in train.py, and start train
    cd Attention_ocr.pytorch
    python train.py --trainlist ./data/ch_train.txt --vallist ./data/ch_test.txt

    then you can see in the terminel as follow: attentionocr there uses the decoderV2 model for decoder.

    The previous version

    git checkout AttentionOcrV1

    Reference

    1. crnn.pytorch
    2. Attention-OCR
    3. Seq2Seq-PyTorch
    4. caffe_ocr

    TO DO

    • change LSTM to Conv1D, it can greatly accelerate the inference
    • change the cnn bone model with inception net, densenet
    • realize the decoder with transformer model

    项目简介

    🚀 Github 镜像仓库 🚀

    源项目地址

    https://github.com/andy-zhujunwen/attention_ocr.pytorch

    发行版本

    当前项目没有发行版本

    贡献者 8

    开发语言

    • Python 100.0 %