CTPN_CRNN_ChineseOCR_PyQt5

CTPN and CRNN based Chinese OCR, developed with PyQt5

Examples

• Hello guys, hope you are doing fine these days !😄
• In this repositories, i created a PyQt5 Application to do some Chinese OCR job which is based on CTPN and CRNN
• Here is the result, hope you enjoy it
• However, the CRNN model doesn't work well(Terrible at some situation😥). You will find it out when you try
• 

Requirements

• PyQt5
• torch
• torchvision

How to use it

• I'm sorry guys, i haven't find a way to deploy it, and you have to run it in your compiler 🙇‍♂️(VScode, pycharm or...)

Build Environment

cd CTPN_lib/bbox
python setup.py build

• a file named build will be generated, put bbox.cp37-win_amd64.pyd and nms.cp37-win_amd64.pyd CTPN_lib/bbox

download pretrained weights

• The weights are larger than the uploading limit(25M below😅). Download them using BaiduYun
• Put them in CTPN_weights CTPN weights(提取码:r88t)
• Put them in CRNN_weights CRNN weights(提取码:tnts)

Choose which model to use

• In inferrence.py, from line 27 to line 32
• argument: crnn_weights: the file location of crnn weigth downloaded in the previous step
• argument: ctpn_basemodel:choose a ctpb backbone: vgg16, resnet50, shufflenet_v2_x1_0, mobilenet_v3_large, mobilenet_v3_small
• argument: ctpn_weights:corresponding ctpn weights with ctpn_base model downloaded in the previous step

Run main_gui.py

• if you run the .py file succesfully, it should look like this
• 
• Then, you need to push the initialize button to load the model, after that, just wait the Finished sign appers in the right.
• 
• Finally, load the image with Load Image button and press Detect
• 

For more issue, contact me

• Email Address mountchicken@outlook.com