图像处理
佐倉
皆さん、こんにちは。私は桜と申しますが、故郷は河南です。ソフトウェアが大好きだなあ。時間があれば一緒に勉強しましょう!よろしく、お願い致します!
展开
-
Linux Mediapipe安装
下载Mediapipe安装bazel升级gcc到gcc8安装opencvhttps://files.pythonhosted.org/packages/1b/82/f4f0e73505fb3a54f974c926bce9f9fe250e50eb57d886daf9042cec0d9b/opencv-contrib-python-4.5.4.58.tar.gzpip3 install -i https://pypi.tuna.tsinghua.edu.cn/simple scikit-build参原创 2021-11-08 08:42:30 · 3294 阅读 · 0 评论 -
jetbot 小车配置
小车组装参考,https://www.bilibili.com/video/BV1Qb4y1C7cj?from=search&seid=3700412571447605611&spm_id_from=333.337.0.0Ubuntu系统安装系统下载地址,https://developer.nvidia.com/embedded/downloads系统安装参考,https://blog.csdn.net/qq_36780295/article/details/108449150原创 2021-11-04 15:39:45 · 612 阅读 · 0 评论 -
利用dlib库实现嘴巴位置识别
所需模块安装pip3 install python-opencvpip3 install dlibpip3 install imutils dlib面部关键点信息图我们实现的是嘴巴识别,关键的开始与结束点为49-67。dlib获取嘴巴位置# -*- coding=utf-8 -*-import cv2import dlibimport osfrom imutils import face_utilsshape_detector_path = os.path.join("mod原创 2021-11-01 21:23:55 · 1411 阅读 · 0 评论 -
face_recognition实现人脸相似度比较
face_recognitionface_recognition 是github上一个非常有名气的人脸识别开源工具包,我们可以通过以下指令安装到python环境内$ pip install face_recognition加载人脸图片image = face_recognition.load_image_file(src)获取图片人脸定位[(top,right,bottom,left )]face_locations = face_recognition.face_locations(i.原创 2021-10-29 15:33:13 · 8403 阅读 · 1 评论 -
训练模型并重建人脸
https://blog.csdn.net/weixin_41943311/article/details/102370766转载 2021-10-26 10:27:19 · 146 阅读 · 0 评论 -
ResNet50网络
自从ResNet在2015年被提出,在ImageNet比赛classification任务上获得第一名,因为它“简单与实用”并存,之后很多方法都建立在ResNet50或者ResNet101的基础上完成的,检测,分割,识别等领域都纷纷使用ResNet,Alpha zero也使用了ResNet,所以可见ResNet确实很好用。Resnets50 网络结构示意图ResNet50和ResNet101对比...原创 2021-10-26 09:00:24 · 693 阅读 · 0 评论 -
cv2嵌入PyQt显示图像与视频
import sys, osfrom PyQt5 import QtCore, QtWidgets, QtGuifrom PyQt5.QtWidgets import *from PyQt5.QtGui import *import cv2class CV2_PYQT_Window(QDialog): def __init__(self,src,target="video",parent=None): super(CV2_PYQT_Window,self).__init原创 2021-10-12 10:49:46 · 3030 阅读 · 0 评论 -
图片处理后进行Kmeans聚类
import randomfrom sklearn import datasetsimport numpy as npimport matplotlib.pyplot as pltfrom mpl_toolkits.mplot3d import Axes3Dimport cv2import numpy as npimport matplotlib.pyplot as pltimport pandas as pdfrom sklearn.cluster import KMeansfrom原创 2021-10-01 11:59:50 · 204 阅读 · 0 评论 -
Python 对图片RGB进行分离与重组
图片组成[[[245 227 220][245 226 219][245 227 220]…[ 82 123 78][ 83 125 78][ 88 131 80]],[[…]],…]分离RGB值def writeFile(X,filename): if filename!="": with open(filename,"w")as f: json.dump(X,f,indent=4)def getMatrixRGB(原创 2021-10-01 11:06:26 · 1003 阅读 · 0 评论 -
目标检测图片相似度比较
目标检测区域写入文件# -*- coding=utf-8 -*-import numpy as np import cv2import osimport csvimport randomimport jsonimport pickleimport timeclass pickle_pic_datasets(object): def __init__(self): self.data = None def get_pic_list(self,filena原创 2021-06-29 17:18:44 · 443 阅读 · 0 评论 -
Python 五种图片相似度比较方法
均值哈希算法# -*- coding: utf-8 -*-import cv2import numpy as np# Hash值对比def cmpHash(hash1, hash2,shape=(10,10)): n = 0 # hash长度不同则返回-1代表传参出错 if len(hash1)!=len(hash2): return -1 # 遍历判断 for i in range(len(hash1)): # 相等则n计原创 2021-06-28 16:04:23 · 24836 阅读 · 9 评论 -
【人体关键点定位】mediapipe_人脸关键点定位(四)
函数式import cv2import mediapipe as mpimport timeimport osimport randomdef video_ope(file): ctime = 0 ptime = 0 Frame = 0 switch = True cap = cv2.VideoCapture(file) myface_mesh = mp.solutions.face_mesh myDraw = mp.solutions.原创 2021-06-08 10:08:05 · 2762 阅读 · 0 评论 -
【人体关键点定位】mediapipe_人脸识别(三)
函数式import cv2import mediapipe as mpimport timeimport osimport randomdef video_ope(file): Frame = 0 switch = True cap = cv2.VideoCapture(file) myface_detection = mp.solutions.face_detection myDraw = mp.solutions.drawing_utils原创 2021-06-08 10:01:53 · 1345 阅读 · 8 评论 -
【人体关键点定位】mediapipe_身体定位(二)
人体关键点图例函数式import cv2import mediapipe as mpimport timeimport osimport randomdef video_ope(file): frame = 0 switch = True cap = cv2.VideoCapture(file) mypose= mp.solutions.pose pose = mypose.Pose() myDraw = mp.solutions.drawi原创 2021-06-04 14:30:17 · 2882 阅读 · 0 评论 -
【人体关键点定位】mediapipe_手部定位(一)
参考:https://google.github.io/mediapipe/solutions/hands函数式import cv2import mediapipe as mpimport timeimport osimport randomdef video_ope(file): switch = True cap = cv2.VideoCapture(file) myhands= mp.solutions.hands hands = myhands.Ha.原创 2021-06-04 14:00:29 · 1145 阅读 · 0 评论 -
目标检测_利用tensorflow2官方案例-自定义训练目标进行人眼识别
我的数据集:说明:我的数据集一共1035张,并非通过手动标记得到,因此不包含xml文件包含:图片文件,train.csv,test.csv。(图片位置,标记位置,目标名称等)说明:数据集本人通过制作而成,只供学习使用,不得私自滥用。链接:https://pan.baidu.com/s/18FdxUHiLnD1B52Jpt4QC6g提取码:zw6b效果展示:单眼模型效果双眼模型效果工程目录:...原创 2021-05-31 11:58:47 · 1639 阅读 · 12 评论 -
darknet训练自定义模型
下载标注工具,标记darknet下载https://github.com/AlexeyAB/darknetvs2017生成darknet.exehttps://blog.csdn.net/qq_38641985/article/details/117227981配置环境变量生成txt位置文件# coding: utf-8import globimport xml.etree.ElementTree as ET # 标记类名class_names = ['tortoise',原创 2021-05-26 17:33:14 · 265 阅读 · 0 评论 -
VS2017配置darknet环境配置【目标检测】
项目下载https://github.com/AlexeyAB/darknet进入工程目录,darknet\build\darknet找到darknet.sln,vs2017打开此处改为修改属性页cuda位置opencv配置包含目录用属性下的VC++目录,右边会有包含目录和库目录,点击包含目录,添加以下三条路径,其实这些都是刚才OpenCV相关解压文件所在的目录opencv\build\includeopencv\build\include\opencv2\coreop原创 2021-05-24 18:19:24 · 591 阅读 · 2 评论 -
利用yolov4进行手写数字识别
识别效果训练20000次数据集下载,https://download.csdn.net/download/qq_38641985/18963935图片标记参考https://github.com/theAIGuysCode/tensorflow-yolov4-tflite修改配置#! /usr/bin/env python# coding=utf-8from easydict import EasyDict as edict__C原创 2021-05-22 11:19:12 · 3660 阅读 · 7 评论 -
yolov4初探学习_预测
地址https://github.com/theAIGuysCode/tensorflow-yolov4-tflite配置环境name: yolov4-gpudependencies:python==3.7pipmatplotlibopencvcudnncudatoolkit==10.1.243pip:tensorflow-gpu==2.3.0rc0opencv-python==4.1.1.26lxmltqdmabsl-pyeasydictpillow我的t原创 2021-05-17 15:25:57 · 177 阅读 · 0 评论 -
手写数据集自动生成
手写数字生成import randomimport osfrom PIL import Image, ImageDraw, ImageFontrandom.seed(3)path_img = "data/"def generate_single(): im_50_blank = Image.new('RGB', (50, 50), (255, 255, 255)) draw = ImageDraw.Draw(im_50_blank) num = str(random.r原创 2021-05-15 22:59:22 · 147 阅读 · 0 评论 -
利用官方案例进行训练自己的目标检测模型
准备工作请参考上篇https://github.com/tensorflow/models标注工具https://github.com/tzutalin/labelImg安装对应模块,调试环境即可利用标注工具制作自己的数据集,并生成xml文件生成csv# -*- coding:utf-8 -*-import osimport globimport pandas as pdimport xml.etree.ElementTree as ETimport randomdef x原创 2021-05-14 09:14:01 · 277 阅读 · 1 评论 -
官方案例进行目标检测_SSD
第一部下载对应源码: https://github.com/tensorflow/models建立工程目录复制models-master\research\object_detection到工程之中下载模型下载对应模型:https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf1_detection_zoo.md这里用的是第一个建立相应目录models放入解压后的模型目录t原创 2021-05-12 11:31:13 · 240 阅读 · 0 评论 -
opencv目标追踪
# -*- coding:utf-8 -*-import numpy as npimport argparseimport cv2points=[]opencv_object_trckers = { "csrt": cv2.TrackerCSRT_create, "kcf": cv2.TrackerKCF_create, "boosting": cv2.TrackerBoosting_create, "mil": cv2.TrackerMIL_create,原创 2021-05-10 08:26:21 · 175 阅读 · 0 评论 -
基于SSD模型的目标追踪
调用SSD模型# -*- coding: utf-8 -*-from tensorflow.python.keras.preprocessing.image import img_to_arrayfrom tensorflow.python.keras.applications.imagenet_utils import preprocess_inputimport cv2 as cvimport numpy as npfrom nets.ssd_net import SSD300from原创 2021-04-27 19:29:30 · 500 阅读 · 0 评论 -
目标检测数据集预处理
数据集:http://host.robots.ox.ac.uk:8080/pascal/VOC/voc2011/index.html说明:本文使用2012工程目录datasets|_____commodity______Annotations||__________JPEGImages数据预处理from xml.etree import ElementTree as ET from collections import OrderedDict as Dictimport num原创 2021-04-25 11:17:56 · 279 阅读 · 0 评论 -
Java opencv读取图片
package open_txt;import org.opencv.core.Core;import org.opencv.core.Core.MinMaxLocResult;import org.opencv.core.Mat;import org.opencv.core.Point;import org.opencv.core.Scalar;import org.opencv.imgcodecs.Imgcodecs;import org.opencv.imgproc.Imgproc;i原创 2021-04-22 21:11:09 · 632 阅读 · 0 评论 -
java opencv配置
https://blog.csdn.net/dongd_70/article/details/82690927转载 2021-04-22 21:07:53 · 190 阅读 · 0 评论 -
C++ opencv简单边缘检测
#include <opencv2/core/core.hpp> #include <opencv2/highgui/highgui.hpp> #include <opencv2/imgproc/imgproc.hpp>#include <iostream>using namespace std;using namespace cv;void display(Mat img){ imshow("display", img); wait原创 2021-04-16 16:47:04 · 325 阅读 · 0 评论 -
vs2017 opecv配置
https://blog.csdn.net/qq_41175905/article/details/80560429转载 2021-04-16 16:12:03 · 68 阅读 · 0 评论 -
基于dlib的眼睛检测
# -*- coding=utf-8 -*-import numpy as np import cv2import dlibfrom scipy.spatial import distanceimport osfrom imutils import face_utilsimg_eye = cv2.imread("eye.png")switch = FalseVECTOR_SIZE = 3def queue_in(queue, data): ret = None if l原创 2021-04-16 12:51:07 · 1154 阅读 · 0 评论 -
光流估计
# -*- coding:utf-8 -*-import cv2import numpy as npdef video_operate(): capture = cv2.VideoCapture("test.avi") feature_param = dict(maxCorners =100,qualityLevel = 0.3,minDistance= 7) lk_param = dict(winSize=(15,15),maxLevel=2) color_ ..原创 2021-04-13 15:44:32 · 94 阅读 · 0 评论 -
背景建模
# -*- coding:utf-8 -*-import cv2import numpy as npdef video_operate(): capture = cv2.VideoCapture("test.avi") kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE,(3,3)) fbgb = cv2.createBackgroundSubtractorMOG2()# 混合高斯模型建模 while Tru...原创 2021-04-13 10:20:02 · 110 阅读 · 0 评论 -
傅里叶变换
低通滤波# -*- coding:utf-8 -*-import numpy as npimport cv2from matplotlib import pyplot as pltdef display(img): cv2.imshow("img",img) cv2.waitKey() cv2.destroyAllWindows()def operate(img): img_float32 = np.float32(img) dft = cv2.dft(img_float32,f..原创 2021-04-09 20:28:47 · 77 阅读 · 0 评论 -
直方图操作
图像直方图def img_dislay(img): hist = cv2.calcHist([img],[0],None,[256],[0,256]) plt.hist(img.ravel(),256) plt.show()图像rgb信息def rgb_display(img):#图像输入必须rgb类型 color = ('b','g','r') for i,col in enumerate(color): histr = cv2.calc原创 2021-04-09 20:22:03 · 87 阅读 · 0 评论 -
图像梯度
Sobel算子def sobel_(src): sobel_x = cv2.Sobel(src,cv2.CV_64F,1,0,ksize=3) sobel_y = cv2.Sobel(src,cv2.CV_64F,0,1,ksize=3) sobel_x = cv2.convertScaleAbs(sobel_x) sobel_y = cv2.convertScaleAbs(sobel_y) sobel_xy = cv2.addWeighted(sobel_x,原创 2021-04-07 21:23:14 · 86 阅读 · 1 评论 -
形态学操作
显示图像def display(img): cv2.imshow("img",img) cv2.waitKey() cv2.destroyAllWindows()腐蚀操作def erode_(img): kernel = np.ones((5,5),np.uint8) erodesion = cv2.erode(img,kernel,iterations=4) display(erodesion)膨胀操作def dilate_(img): kernel = np.ones((原创 2021-04-06 16:43:02 · 127 阅读 · 0 评论 -
常用滤波器
# -*- coding: utf-8 -*-import cv2 as cvimport numpy as npdef display_img(img): cv.imshow('image',img) cv.waitKey(0) cv.destroyAllWindows()def average_(filepath):#均值滤波,适用于去除通过扫描得到的图像中的颗粒噪声 img = cv.imread(filepath) blur = cv.blur(i原创 2021-04-05 10:44:58 · 194 阅读 · 0 评论 -
二值化处理
import cv2 as cvimport matplotlib.pyplot as pltthresh_type = [ cv.THRESH_BINARY, cv.THRESH_BINARY_INV, cv.THRESH_TRUNC, cv.THRESH_TOZERO, cv.THRESH_TOZERO_INV ]title = [ "ORIGNAL", "BINARY", ".原创 2021-04-05 10:14:04 · 612 阅读 · 0 评论 -
目标跟踪
目标跟踪源码# -*- ending:utf-8 -*-import dlibimport numpy as npimport argparseimport cv2points=[]def show_info(frame,tracking_status): pos1 = (10,20) pos2 = (10,40) pos3 = (10,60) info1 = "click selct an area,start tracking" info2 = "'原创 2021-04-03 14:25:24 · 86 阅读 · 3 评论