Hands-on OCR (1)

文本检测：DBNet

环境：paddle

import matplotlib
from Cython import inline
from paddleocr import PaddleOCR

ocr=PaddleOCR()
#修改图片路径
img_path='D:/computervision/ocr/ppocr_img/imgs/12.jpg'

result =ocr.ocr(img_path,rec=False)
print(f'the predict as follow')
print(result)

import numpy as np
import cv2
import matplotlib.pyplot as plt

image = cv2.imread(img_path)
#The following sentence seems optional and does not affect the structure
boxes=[line[0] for line in result]

for box in result:
    box=np.reshape(np.array(box),[-1,1,2]).astype(np.int64)
    image=cv2.polylines(np.array(image),[box],True,(255,0,0),2)

plt.figure(figsize=(10,10))
#The code has been modified here
plt.imshow(image)
plt.show()

paddle

1. 核心代码

声明PaddleOCR类：

ocr=PaddleOCR()

2. 执行预测：

result =ocr.ocr(img_path,rec=False)

• 运行结果
  

- Please explain other libraries

1. reshape改变数组形状

• 参数

（1）传入数组
（2）排序方式

A shape dimension can be-1.在这种情况下,The value is inferred from the length and remaining dimensions of the array.

Example from withcsdnanother blogger

（3）order
‘C’ means C order, ‘F’ means Fortran order
orderIn fact, I don't really understand it to be honest,I'll add it when I figure it out

2)astype
强制转化数据类型

3）figure
figure(num=None, figsize=None, dpi=None, facecolor=None, edgecolor=None, frameon=True)

• 作用：My understanding is kind of likejavafx的scence.
• 参数

num:图像编号或名称,数字为编号 ,字符串为名称
figsize:指定figure的宽和高,单位为英寸;
dpi参数：指定绘图对象的分辨率,即每英寸多少个像素,缺省值为80
facecolor:背景颜色
edgecolor:边框颜色
frameon:是否显示边框

DB文本检测模型

import paddle
import os
import sys
import importlib

# paddleocr_path='D:/computervision/ocr/PaddleOCR-release-2.5/PaddleOCR-release-2.5/ppocr'
# sys.path.append( paddleocr_path )
# os.chdir(paddleocr_path)
# print("当前工作目录"+os.getcwd())


#import into the path
from ppocr.modeling.backbones.det_mobilenet_v3 import MobileNetV3
fake_input=paddle.randn([1,3,640,640],dtype='float32')

model_backbone=MobileNetV3()
model_backbone.eval()

outs=model_backbone(fake_input)
print(model_backbone)

for idx,out in enumerate(outs):
    print("the index",idx,"shape:",out.shape)

ppocrThe path is different from the current file working path and was not found at first,I tried many methods to no avail,used in the article第2种办法to read the file