Python Computer Vision

One application direction of machine learning is how to make machines understand images. This includes the identification, tracking and measurement of objects in the image.

What to do -driverless cars, face recognition, license plate recognition gesture recognition (game orientation)

PIL Static Library

OpenCV Dynamic Library

ImportPandas as PDImportNumPy as NP fromPILImportImagetrain=pd.read_csv ('Train.csv') forInd,rowinchTrain.iloc[1:10].iterrows ():#the first line is the labelI=Row[0] Arr=np.array (row[1:],dtype=np.uint8) arr.resize ((28,28)) Im=Image.fromarray (arr)#IND is the first image, I means that the image is a few    #if there is no ' train_pics ' This folder will be errorIm.save ('./train_pics/%s-%s.png'% (Ind,i))

OpenCV Video input and output-this program is not running because of environmental configuration issues:

################# camera input, output ##################defOnmouse (event, x, Y, Flags, param):#Event:    #cv_event_mousemove 0 Slide    #Cv_event_lbuttondown 1 left button click    #Cv_event_rbuttondown 2 Right-click    #Cv_event_mbuttondown 3 middle button click    #cv_event_lbuttonup 4 left button release    #Cv_event_rbuttonup 5 Right-click Release    #cv_event_mbuttonup 6 middle button Release    #cv_event_lbuttondblclk 7 left button double-click    #CV_EVENT_RBUTTONDBLCLK 8 Right-click    #CV_EVENT_MBUTTONDBLCLK 9 middle button Double click    #x, y for mouse click position    #Flags:    #Cv_event_flag_lbutton 1 left button drag    #Cv_event_flag_rbutton 2 Right-click Drag    #Cv_event_flag_mbutton 4 middle button Drag    #Cv_event_flag_ctrlkey 8 (8~15) press CTRL to leave events    #Cv_event_flag_shiftkey 16 (16~31) press shift to leave an event    #Cv_event_flag_altkey 32 (32~39) press ALT to leave an event        #param: Custom numbering    Globalclickedifevent = =cv2.cv.CV_EVENT_LBUTTONUP:clicked=trueclicked=False#Read camera inputCameracapture =Cv2. Videocapture (0) Cv2.namedwindow ('Mycamera')#Bind Mouse CallbackCv2.setmousecallback ('Mycamera', Onmouse)PrintU'Click the window or press any key to exit.'Success, Frame=Cameracapture.read () whileCv2.waitkey (1) = =-1 and  notclicked:ifFrame is  notNone:cv2.imshow ('Mycamera', frame) success, frame=Cameracapture.read () Cv2.destroywindow ('Mycamera')

First, Haar cascade classifier

Harr Cascade classifier =harr-like feature detection +adaboost

How to make a few weak classifiers into a strong classifier.

#-*-coding:utf-8-*-#Comment by HeibankeImportCv2#instantiating a classifier#The argument to the instantiated function is an XML file, which is a strong classifier that trains the milk well.Face_cascade = Cv2. Cascadeclassifier ('./xml/haarcascade_frontalface.xml') Eye_cascade= Cv2. Cascadeclassifier ('./xml/haarcascade_eye.xml')#Open Imageimg = Cv2.imread ('./pics/test_faces.jpg') Gray=Cv2.cvtcolor (IMG, Cv2. Color_bgr2gray)#Gray = cv2.imread (' test1.jpg ', Cv2. Cv_load_image_grayscale)#Detectmultiscale Parameter Explanation#Gray , grayscale image for detection#1.2:scale_factor The scale factor of the search window in successive scans of the previous and posterior two times. --that is, when searching, zoom in on the window proportionally.#For example, 1.1 refers to expanding the search window by 10%. #2:min_neighbors The minimum number of adjacent rectangles that comprise the detection target (Default-1). --only judging the adjacent rectangle is sometimes judged as a human face, if it is not, then it will not be treated as a human face .#if the number of small rectangles that comprise the detection target and less than min_neighbors-1 are excluded. #if Min_neighbors is 0, the function returns all the checked candidate rectangles without any action. --We choose 2, which is the rectangle that selects all the adjacent rectangles that are faces.faces = Face_cascade.detectmultiscale (Gray, 1.2, 2) for(X,Y,W,H)inchfaces:#Cv2.rectangle Parameter Explanation    #(x, y) is the upper-left corner of the rectangle    #(X+W,Y+H) is the lower right corner of the rectangle    #(255,0,0) is the RGB color of the rectangle, which is red    #2, is the line width of the drawing rectangleCv2.rectangle (IMG, (x, y), (x+w,y+h), (255,0,0), 2) Roi_gray= Gray[y:y+h, x:x+W] Roi_color= Img[y:y+h, x:x+W]#detection of the eyes based on human face detectionEyes =Eye_cascade.detectmultiscale (Roi_gray) for(Ex,ey,ew,eh)inchEyes:cv2.rectangle (Roi_color, (Ex,ey), (ex+ew,ey+eh), (0,255,0), 2) Cv2.imshow ('img', IMG) k=cv2.waitkey (0)ifK==27: Cv2.destroywindow ('Test')

Python Computer Vision