PaddleOCR User Guide

First install paddle, then install paddleocr

pip install "paddleocr>=2.0.1"

Identify images

from paddleocr import PaddleOCR, draw_ocrfrom PIL import Imageif __name__ == '__main__': ocr = PaddleOCR(use_angle_cls=True, lang='ch') img_path = 'demo/demo_kie.jpeg' result = ocr.ocr(img_path, cls=True) for line in result: print(line) image = Image.open(img_path).convert('RGB') boxes = [line[0] for line in result] txts = [line[1][0] for line in result] scores = [line[1][1] for line in result] im_show = draw_ocr(image, boxes, txts, scores, font_path='data/chineseocr/labels/font.TTF') im_show= Image.fromarray(im_show) im_show.save('output/result5.jpg')

The lang in the PaddleOCR(use_angle_cls=True, lang='ch') here can be many languages, such as `ch`, `en`, `fr`, `german`, `korean`, `japan`.

This includes both text detection and text recognition. The general results are as follows

But if it is a relatively simple text, such as

At this time, we only need to identify, no need to detect

from paddleocr import PaddleOCR, draw_ocrif __name__ == '__main__': ocr = PaddleOCR(use_angle_cls=True, lang='en') img_path = 'demo/demo_text_recog.jpg' result = ocr.ocr(img_path,cls=True, det=False) for line in result: print(line)

Running Results (Partial)

('STAR', 0.8838256597518921)

Paddle OCR framework download address: GitHub - PaddlePaddle/PaddleOCR: Awesome multilingual OCR toolkits based on PaddlePaddle (practical ultra lightweight OCR system, support 80+ languages ​​recognition,provide data annotation and synthesis tools, support training and deployment among server, mobile, embedded and IoT devices)

Model training

Here is still taking Kaggle verification code text recognition as an example. The data set format of PaddleOCR is somewhat different from that of MMOCR. It needs to put the images of the training data set and the test data set in two differentin the folder.The general style is as follows

Since they were all put together before, write a script to separate them

import shutilif __name__ == '__main__': with open('data/toy_dataset/test_label.txt', 'r') as f: for line in f: filename = line.split(' ')[0] shutil.move('data/toy_dataset/train/' + filename, 'data/toy_dataset/test/' + filename)

In addition, its tag files are separated by tabs \t, while in MMOCR they are separated by spaces.

2wc38.png 2wc38y5n6d.png y5n6dmen4f.png men4f57b27.png 57b27x3deb.png x3deb

Modify the configs/rec/rec_icdar15_train.yml file in the PaddleOCR main directory. Of course, this is only one of the recognition frameworks. Let's take this as an example, and the modified parts are as follows

Train: dataset: name: SimpleDataSet# data_dir: ./train_data/ic15_data/ data_dir: ./data/toy_dataset/train/# label_file_list: ["./train_data/ic15_data/rec_gt_train.txt"] label_file_list:["./data/toy_dataset/train_label.txt"] transforms: - DecodeImage: # load image img_mode: BGR channel_first: False - CTCLabelEncode: # Class handling label - RecResizeImg: image_shape: [3, 32, 100] # Chinese[3, 32, 320] - KeepKeys: keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order loader: shuffle: True batch_size_per_card: 256 drop_last: True num_workers: 8 use_shared_memory: FalseEval: dataset: name: SimpleDataSet# data_dir: ./train_data/ic15_data data_dir: ./data/toy_dataset/test/# label_file_list: ["./train_data/ic15_data/rec_gt_test.txt"] label_file_list: ["./data/toy_dataset/test_label.txt"] transforms: - DecodeImage: # load image img_mode: BGR channel_first: False - CTCLabelEncode: # Class handling label - RecResizeImg: image_shape: [3, 32, 100] - KeepKeys: keep_keys: ['image', 'label', 'length'] # dataloader will return list in thisorder loader: shuffle: False drop_last: False batch_size_per_card: 256 num_workers: 4 use_shared_memory: False

Copy train.py in the tools folder to the PaddleOCR main folder and add parameters

--config=configs/rec/rec_icdar15_train.yml

Run, start training.