秀呀秀的博客

【代码】python-旋转矩形根据四个点的矩形框顺序进行方向矫正并裁剪。

更新完成之后，重新训练，没有报错了！

针对yolov5 pt转onnx报错进行修改，并根据转换成功的onnx进行预测输出 修改： 去掉Detect层，改为转换后的处理，对三个输出层处理（export grid=True)转换后的onnx的输出少了直接通过detect层输出的[1,6300,9] ： 对输出的onnx进行detect层输出并预测（参考链接：基于onnxruntime的YOLOv5单张图片检测实现_wxplol的博客-CSDN博客_onnx yolov5）：......

在tensorflow2中对savedmodel文件夹下文件转换成tflite

tensorflow2中对savedmodel文件夹下保存的文件进行frozen_graph

import cv2import osimport numpy as np#color: QR(1):128 0 0 MINIQR(2):0 128 0 DM(3):128 128 0 PDF(4):0 0 128img_path='/data1/gyx/QR/multiyolov5_detect_seg/data/customdata/convert_tools/qr/segvoc/SegmentationClassPNG/'out_img='/data1/gyx/.

使用之前注意如下三点：1.确定原始图像存储路径以及掩码文件存储路径2.路径下的图像格式要保持一致 比如都是PNG （不然生成不了，检测不到图片）3.image和label的mode格式最好都是RGB （不然会报错如：ValueError: image has wrong mode）#######################数据增强工具##################import Augmentorp = Augmentor.Pipeline("/data1/gyx/QR/mul..

如果conda环境没问题，直接升级就好了：conda install python=3.7如果出现如下错误：CondaHTTPError: HTTP 000 CONNECTION FAILED for url <https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge/noarch/repodata.json>试一试验证：conda config --set ssl_verify false如果还不行，则需...

如果安装doctr出现如下error:The conflict is caused by: python-doctr 0.4.1 depends on mplcursors>=0.3 python-doctr 0.4.0 depends on mplcursors>=0.3 python-doctr 0.3.1 depends on mplcursors>=0.3 python-doctr 0.3.0 depends on mplcursors>...

##### 根据同文件夹一个xml，生成其他图像的xml #####import osimport xml.etree.ElementTree as ETfrom xml.etree.ElementTree import parse, Elementimport shutil pcb_list=os.listdir('./')print(pcb_list)for folder in pcb_list: if folder.endswith('py'): pass .

############### 判断字体中是否都包含字典中的字符 ################from fontTools.ttLib import TTFontimport osfrom shutil import copyfilefrom fontTools.ttLib.sfnt import DirectoryEntrywith open('/data/git/ocr-platform/statistic/TextGenerator/TextGenerator/process/dict.

#使用多线程运行from multiprocessing import Poolfrom tqdm import tqdmif __name__=='__main__': start_num=0 numbers=2000 thread_count=4 p = Pool(thread_count) for _ in tqdm( p.imap_unordered( QRDataGenerator.generate_from_.

import randomfrom random import randintimport osimport reimport stringdict_num='0123456789'dict_num_letter='0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ'dict_num_letter_all='0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz'dict_symbol='!@#$%.

'''单通道->三通道'''import osimport cv2import numpy as npimport PIL.Image as Imageimport osos.environ['CUDA_VISIBLE_DEVICES'] = '2'img_path='/home/gyx/QR/qr_detect_model/dataset/images_all_channel_1/'save_img_path='/home/gyx/QR/qr_detect_model/data.

删除S 之后的所有字符用：s.*$删除S 之前的所有字符用：^.*s删除前：想要删除斜杠‘/’之前的字符：^.*/完成后，删除成功：

from PIL import Imageimport osimport cv2import timeimport numpy as npsave_img_path='result_tflite/'img_path="C:/Users/DELL/Desktop/20210416/"def shot(img,img_name,boxes):#应用于predict_det.py中,通过dt_boxes中获得的四个坐标点,裁剪出图像 index=0 img_resized=im.

遍历目录下包括子文件夹等所有下属的文件以及文件夹，查找空文件以及空文件夹并删除：def Clean_empty(path): """ 遍历文件下所有子文件夹以及子文件，清理空文件夹和空文件 path:文件路径 """ for (dirpath,dirnames,filenames) in os.walk(path): for filename in filenames: file_folder=dirpath+'

ImportError: No module named 'model'ImportError: No module named 'xxx'看了一眼目录结构是有model的，为什么from model.densenet_crnn import DenseNetImportError: No module named 'model' 还是显示不了呢：原来是需要把模块路径放到环境变量中作为全局变量sys.path.append('F:/OCR/code/ocr_densenet_te

1.一开始使用的是下面这种旋转方法：def rotate(pic): img = cv2.imread(pic) # img = Image.open(pic) # img = np.array(img) height, width = img.shape[:2] degree = 5 # 旋转后的尺寸 heightNew = int(width * fabs(sin(radians(degree))) + height * fabs(cos(radians(degree))))

如果想同时在两个列表中取值的话，可以使用zip:for i,j in zip([0,1,2],[0,1,2]) : print(i,j)结果为：0 01 12 2来进行高阶的写法：for image,mask in zip([image0,image1,image2],[mask0,mask1,mask2]): rotated_img = image.rotate( skewing_angle i.

import cv2import numpy as npimport pyzbar.pyzbar as pyzbarfrom skimage import iourl = "https://d33wubrfki0l68.cloudfront.net/19994bca1018ba79dd954bca9edc627dcd1f1dc0/d9ac1/wp-content/uploads/2018/11/19.jpg"image = io.imread(url)#也可以选择自己的qr码图像# imag.

去除关键词前的txt文档key_word.txt和除去相同关键词新建的txt文档：key_word_new.txt# _*_ coding: utf-8 _*___author__ = 'gyx'__date__ = '2020/3/23 14:05'import jiebaimport pandas as pdimport osfrom collections import CountertxtPath = 'key_word.txt'txtPath_new =

'''Extract panel :kmeans聚类'''import osimport cv2import numpy as npimport mathdef panelAbstract(srcImage): # read pic shape print(srcImage.shape) imgHeight,imgWidth = srcImage.sha...

def image2label(path,size_): w = size_[0] h = size_[1] label_im=cv2.imread(path) new_array = cv2.resize(label_im, (w, h), interpolation=cv2.INTER_CUBIC) data=np.array(new_arr...

读取weibo.xlsx文件：import pandas as pddf = pd.read_excel("weibo.xlsx") worksheet= df.values for file in worksheet: print('file',file) //表格每一行 print(file[0]) //表格每一行的第一列结果如下：...

错误如下：error: json.decoder.JSONDecodeError: Unterminated string starting at: line 1 column 188 (char 187)搜了半天来查找原因，没有解决问题，最后发现竟然是文档最后末尾缺少空行，????添加空行：...

Python2.==>转换成Python3.error：ImportError: No module named 'ConfigParser'import configparsererror：ImportError: No module named 'cPickle'import pickle

如果是列表查找相同元素的个数，可以转换成字典，返回其values的值这里需要插入 counter 包：from collections import Counter举例说明：class_1_list=["喜","怒","哀","乐","乐"]class_1_dict = dict(Counter([k for k in class_1_list]))for k, v ...

利用字典来替换switch case的写法：def polar_chs(polar): switch_case = { 0: 0, 1: 1, 2: -1, 3: 0 } return switch_case.get(polar, 'wrong polar val...

单个json格式'''class.txt{'PA':'乐','PE':'乐','PD':'好','PH':'好','PG':'好','PB':'好'}'''import osimport jsonpath_file = os.path.dirname(os.path.abspath(__file__))class_path = os.path.join(path_file...

jpg格式:是有损图片压缩，可以使用最少的磁盘空间得到质量较好的图片png格式:不属于压缩性。png转换成jpg：from PIL import Imageimport cv2 as cvimport osdef PNG_JPG(PngPath): img = cv.imread(PngPath, 0) w, h = img.shape[::-1] in...

Python3.7 pip安装 dlib报错：ERROR: Command errored out with exit status 1: /home/jovyan/anaconda3/bin/python -u -c 'import sys, setuptools, tokenize; sys.argv[0] = '"'"'/tmp/pip-install-3cbh3fa9/dlib/...

import pandas as pdtrain_sample_metadata = pd.read_json('New_dataset/metadata.json').Ttrain_sample_metadata.head()import csvimport jsondef fk_rl(str): if str == 'FAKE': return 0 els...

Error:UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 0: invalid start byte修改前：import pickleinf = pickle.load(open('Youtube_8M_WILLOW_new/train.pkl'))将打开方式添加‘rb’，修改后：im...

pip国内源清华：https://pypi.tuna.tsinghua.edu.cn/simple阿里云：http://mirrors.aliyun.com/pypi/simple/中国科技大学https://pypi.mirrors.ustc.edu.cn/simple/华中理工大学：http://pypi.hustunique.com/山东理工大学：http://pypi...

interpolation 选项所用的插值方法： INTER_NEAREST 最近邻插值 INTER_LINEAR 双线性插值（默认设置） INTER_AREA 使用像素区域关系进行重采样。 INTER_CUBIC 4x4像素邻域的双三次插值...

# -*- coding: utf-8 -*-__author__ = 'gyx'__date__ = '2019/12/30 17:14'import pandas as pdimport osfrom collections import Counterdef add_keys(political_path,data_key,key): if key not in d...

去除关键词前的txt文档betting_keys和除去相同关键词新建的txt文档：betting_keys_new# _*_ coding: utf-8 _*___author__ = 'gyx'__date__ = '2019/12/20 10:05'import pandas as pdimport osfrom collections import Cou...

import pandas as pdimport os##########需要转换的csv文件###########path_dir = 'E:/1_gyx/Violation_Detect_all/key_words/'csvPath = path_dir + 'betting_keys.csv'if not os.path.exists(csvPath): print...

通过列表的形式进行文本关键词检索，进行博彩类预测，查看程序时间：import jiebaimport pandas as pdimport numpy as npimport osfrom collections import Counter from time import *root_dir = os.path.realpath(__file__ + "/..")bet...