Article Outline
Python pil example 'FacialRecon'
Modules used in program:
import face_recognitionimport pickleimport glob, os, codecs, json, imutils, cv2import numpy as np
python FacialRecon
Python pil example: FacialRecon
from PIL import Image, ImageDraw
from imutils.video import FileVideoStream
from imutils.video import VideoStream
import numpy as np
import glob, os, codecs, json, imutils, cv2
import pickle
import face_recognition
from libs.models import FrameModel
from .models.FaceBundle import FaceBundle
class FaceRecognition:
def __init__(self, storageFolderPath='./storage/',
knownFolderPath='./known/',
unkwonFolderPath='./unknown/',
outputFolderPath='./output',
outputData='outputData.json',
tolerance=0.6):
self.storageFolderPath = storageFolderPath
self.knownFolderPath = knownFolderPath
self.unkwonFolderPath = unkwonFolderPath
self.outputFolderPath = outputFolderPath
self.faceBundleFile = storageFolderPath + 'facebundlefile.obj'
self.csvFile = outputData
self.toleranceRate = tolerance
self.knownFaces = []
if os.path.exists(self.faceBundleFile):
face_file = open(self.faceBundleFile, 'rb')
self.knownFaces = pickle.load(face_file)
face_file.close()
#
# Generic
#
def setKnownFolderPath(self, knownFolderPath):
self.knownFolderPath = knownFolderPath
def setUnknownFolderPath(self, unkwonFolderPath):
self.unkwonFolderPath = unkwonFolderPath
def setOuputFolderPath(self, outputFolderPath):
self.outputFolderPath = outputFolderPath
def setOuputData(self, outputData):
self.outputData = outputData
#
# Private Functions
#
#
# Dada uma imagem, Retorna uma lista com todas as Faces presentes
#
def __parseFaces(self, filePath, image_read=None) -> list:
listFaces: list = []
# Load test image to find faces in
if image_read is None:
image_read = face_recognition.load_image_file(filePath)
# File Identification
filename_w_ext = os.path.basename(filePath)
filename, file_extension = os.path.splitext(filename_w_ext)
# Find faces in test image
# print(("image read:" + image_read))
face_locations = face_recognition.face_locations(image_read)
face_encodings = face_recognition.face_encodings(image_read, face_locations)
face_landmarks = face_recognition.face_landmarks(image_read)
count = 0
for location, encodings, landmarks in zip(face_locations, face_encodings, face_landmarks):
faceBundle = FaceBundle(filename, file_extension, count)
faceBundle.setLocation(location)
faceBundle.setEncodings(encodings)
faceBundle.setLandmarks(landmarks)
listFaces.append(faceBundle)
count += 1
return listFaces
def __hasMatch(self, knownEncodings, faceBundle, debug=True):
matches = face_recognition.compare_faces(knownEncodings, faceBundle.getEncodings(),
tolerance=self.toleranceRate)
if debug:
for i in range(0, len(self.knownFaces)):
if True == matches[i]:
print(faceBundle.getName(), '==', self.knownFaces[i].getName())
return matches
def __drawImage(self, drawInstance, faceBundle, label, drawBox=True, drawLandmarks=True, drawLabel=True):
top, right, bottom, left = faceBundle.getLocation()
# Draw box
if drawBox:
drawInstance.rectangle(((left, top), (right, bottom)), outline=(255, 255, 0))
# Draw Landmarks
if drawLandmarks:
for __, value in faceBundle.getLandmarks().items():
drawInstance.line(value, fill=(27, 189, 230), width=1)
# Draw label
if drawLabel:
text_width, text_height = drawInstance.textsize(label)
drawInstance.rectangle(((left, bottom - text_height - 10), (right, bottom)), fill=(255, 255, 0),
outline=(255, 255, 0))
drawInstance.text((left + 6, bottom - text_height - 5), label, fill=(0, 0, 0))
#
# Public Functions
#
#
# Recorta todas as Faces de uma Imagem e as Salva
#
def pullFaces(self, imagePath, outputExtension='jpg'):
listFaces = self.__parseFaces(imagePath)
image = face_recognition.load_image_file(imagePath)
for face in listFaces:
top, right, bottom, left = face.getLocation()
face_image = image[top:bottom, left:right]
pil_image = Image.fromarray(face_image)
resultPath = "{}/{}.jpg".format(self.outputFolderPath, face.getName())
pil_image.save(resultPath)
print("pullFaces -- {} Faces Found".format(len(listFaces)))
print("pullFaces -- Done")
return listFaces
#
# Known Faces
#
def parseKnownFaces(self, image_extension=''):
if not self.knownFaces:
files_regex = self.knownFolderPath + '*' + image_extension
files = glob.glob(files_regex)
for filename in files:
faceBundleList = self.__parseFaces(filename)
if len(faceBundleList) != 0:
self.knownFaces.append(faceBundleList[0])
else:
print("Encoding Error on", filename)
self.saveKnownFaces()
def addKnownFace(self, filePath):
faceBundleList = self.__parseFaces(filePath)
if len(faceBundleList):
self.knownFaces.append(faceBundleList[0])
self.saveKnownFaces()
return faceBundleList[0].toData()
def findMatches(self, filePath, image_read=None, draw_matches=True, id='') -> list:
faceList: list = []
# Find Faces
faceBundleList = self.__parseFaces(filePath, image_read)
# Prepare to Draw
if draw_matches:
if image_read is None:
image_read = face_recognition.load_image_file(filePath)
pil_image = Image.fromarray(image_read)
draw = ImageDraw.Draw(pil_image)
# Prepare to Match
knownFacesEncoding = []
for knownFace in self.knownFaces:
knownFacesEncoding.append(knownFace.getEncodings())
for faceBundle in faceBundleList:
matches = self.__hasMatch(knownFacesEncoding, faceBundle)
if draw_matches and len(matches) > 0 and True in matches:
first_match_index = matches.index(True)
name = self.knownFaces[first_match_index].getName()
self.__drawImage(draw, faceBundle, name, drawBox=True)
pil_image.save('{}/result_{}.jpg'.format(self.outputFolderPath, id))
faceList.append(faceBundle.toData())
return faceList
def videoRecognition(self, videoPath, fileStream=True, skipFrames=10, showVideo=False, id='') -> list:
actual_frame: FrameModel
frame_list: list = []
# start the video stream thread
print("[INFO] starting video stream thread...")
# to stream from a file
if fileStream:
vs = FileVideoStream(videoPath).start()
else:
vs = VideoStream(src=0).start()
frameCount = 0
vs.thread.setName('Stream Thread')
# loop over frames from the video stream
while True:
# if this is a file video stream, then we need to check if
# there any more frames left in the buffer to process
if fileStream and not vs.more() or vs.stopped:
break
# grab the frame from the threaded video file stream, resize
# it, and convert it to grayscale channels
frame = vs.read()
frame = imutils.resize(frame, width=450)
frameCount += 1
# Recognition
if frameCount % skipFrames == 0:
print('At Frame #{}'.format(frameCount))
face_list = self.findMatches(videoPath, image_read=frame, id='{}_{}'.format(id, frameCount))
actual_frame = FrameModel.FrameModel(videoPath, frameCount)
actual_frame.set_face_list(face_list)
frame_list.append(actual_frame)
# show the frame
if showVideo:
cv2.imshow("Frame", frame)
key = cv2.waitKey(1) & 0xFF
# if the `q` key was pressed, break from the loop
if key == ord("q"):
break
# do a bit of cleanup
if showVideo:
cv2.destroyAllWindows()
vs.stop()
return frame_list
def saveKnownFaces(self):
# dataArray = []
# for faceBundle in self.knownFaces:
# jsonData = faceBundle.toData()
# dataArray.append(jsonData)
# json.dump(dataArray, codecs.open(self.csvFile, 'a', encoding='utf-8'), separators=(',', ':'), sort_keys=True)
try:
face_bundle_file = open(self.faceBundleFile, 'wb')
pickle.dump(self.knownFaces, face_bundle_file)
face_bundle_file.close()
except FileNotFoundError:
print('File {} not Found'.format(self.faceBundleFile))
def parseFromJson(self, json_path):
obj_text = codecs.open(json_path, 'r', encoding='utf-8').read()
b_new = json.loads(obj_text)
print()
# a_new = np.array(b_new)
# def __writeData(self, data):
# with open(self.csvFile, 'a') as csvFile:
# writer = csv.writer(csvFile)
# writer.writerow(data)
# csvFile.close()
Python links
- Learn Python: https://pythonbasics.org/
- Python Tutorial: https://pythonprogramminglanguage.com