Spaces:
Sleeping
Sleeping
File size: 3,672 Bytes
5b0ef18 75ee736 5b0ef18 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 |
import os
os.system('cd ezocr;'
'pip install .; cd ..')
import gradio as gr
import pandas as pd
from PIL import ImageDraw
from easyocrlite import ReaderLite
from PIL import Image
from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks
from modelscope.outputs import OutputKeys
# step 1. orc detection to find boxes
reader = ReaderLite(gpu=True)
# step 2. recognize ocr result according to ocr detection results
ocr_recognize = pipeline(Tasks.ocr_recognition,
model='damo/ofa_ocr-recognition_general_base_zh', model_revision='v1.0.0')
def get_images(img: str, reader: ReaderLite, **kwargs):
results = reader.process(img, **kwargs)
return results
def draw_boxes(image, bounds, color='red', width=4):
draw = ImageDraw.Draw(image)
for i, bound in enumerate(bounds):
p0, p1, p2, p3 = bound
draw.text((p0[0]+5, p0[1]+5), str(i+1), fill=color, align='center')
draw.line([*p0, *p1, *p2, *p3, *p0], fill=color, width=width)
return image
def ofa_ocr_gr():
def ocr_api(img):
results = get_images(img, reader, max_size=4000, text_confidence=0.7, text_threshold=0.4,
link_threshold=0.4, slope_ths=0., add_margin=0.04)
box_list, image_list = zip(*results)
draw_boxes(img, box_list)
ocr_result = []
for i, (box, image) in enumerate(zip(box_list, image_list)):
image = Image.fromarray(image)
result = ocr_recognize(image)[OutputKeys.TEXT][0].replace(" ", "")
ocr_result.append([str(i + 1), result.replace(' ', '')])
result = pd.DataFrame(ocr_result, columns=['Box ID', 'Text'])
return img, result, box_list ##返回box_list 方便微调模型
examples = [
"http://xingchen-data.oss-cn-zhangjiakou.aliyuncs.com/maas/ocr/qiaodaima.png",
"http://xingchen-data.oss-cn-zhangjiakou.aliyuncs.com/maas/ocr/shupai.png",
"http://xingchen-data.oss-cn-zhangjiakou.aliyuncs.com/maas/ocr/ocr_essay.jpg",
"http://xingchen-data.oss-cn-zhangjiakou.aliyuncs.com/maas/ocr/chinese.jpg",
"http://xingchen-data.oss-cn-zhangjiakou.aliyuncs.com/maas/ocr/benpao.jpeg",
"http://xingchen-data.oss-cn-zhangjiakou.aliyuncs.com/maas/ocr/gaidao.jpeg",
]
title = "<h1 align='center'>基于OFA的OCR识别的应用</h1>"
description = '中文OCR体验区，欢迎上传图片，静待检测文字返回~ 相关OCR代码和模型都已在ModelScope开源，支持finetune，欢迎大家在平台上使用！（注：受资源限制，这里只部署了通用OCR模型。）'
ocr_input_image = gr.components.Image(label='图片', type='pil')
ocr_output_image = gr.components.Image(label='图片')
ocr_output_text = gr.components.Dataframe(label='OCR结果', headers=['ID', '文本'])
# ocr_demo = gr.Interface(
# fn=ocr_api,
# inputs=[ocr_input_image],
# outputs=[ocr_output_image, ocr_output_text],
# title=title,
# description=description,
# allow_flagging='never',
# examples=examples,
# examples_per_page=5,
# cache_examples=True
# )
ocr_demo = gr.Interface(
fn=ocr_api,
inputs=[ocr_input_image],
outputs=[ocr_output_image, ocr_output_text],
title=title,
description=description,
allow_flagging='never',
examples=examples,
examples_per_page=5,
cache_examples=False
)
return ocr_demo
if __name__ == "__main__":
ocr_demo = ofa_ocr_gr()
ocr_demo.launch(
share=True,
enable_queue=True,
)
|