def main(args):
mkdirP(args.outputDir)
videos = list(iterVids(args.inputDir))
align = openface.AlignDlib(dlibFacePredictor)
net = openface.TorchNeuralNet(faceDetectionModel, imgDim=96, cuda=False)
# TODO - consider shuffling
processList = [
# "brendan",
# "allen",
# "chris",
# "rohan",
# "justin"
# "eddie",
# "nikita",
# "tyler"
]
for videoObject in videos:
print("=== {} ===".format(videoObject.path))
# if videoObject.cls not in processList:
# continue
# if "2" not in videoObject.name:
# continue
classDir = os.path.join(args.outputDir, videoObject.cls)
mkdirP(classDir)
extract_faces(args, videoObject, classDir, net, align,
videoObject.name, args.multiple)
def align(args, multi):
# Aligning now
start = time.time()
mkdirP(alignedDir)
imgs = list(iterImgs(trainingDir))
# Shuffle so multiple versions can be run at once.
random.shuffle(imgs)
landmarkMap = {
'outerEyesAndNose': openface.AlignDlib.OUTER_EYES_AND_NOSE,
'innerEyesAndBottomLip': openface.AlignDlib.INNER_EYES_AND_BOTTOM_LIP
}
if landmarks not in landmarkMap:
# TODO: Avoid exceptions, find way to silently fail and skip image
raise Exception("Landmarks unrecognized: {}".format(landmarks))
landmarkIndices = landmarkMap[landmarks]
align = openface.AlignDlib(dlibFacePredictor)
nFallbacks = 0
for imgObject in imgs:
print("=== {} ===".format(imgObject.path))
outDir = os.path.join(alignedDir, imgObject.cls)
mkdirP(outDir)
outputPrefix = os.path.join(outDir, imgObject.name)
# TODO: output is still PNG?
imgName = outputPrefix + ".png"
if os.path.isfile(imgName):
if args.verbose:
print(" + Already found, skipping.")
else:
rgb = imgObject.getRGB()
if rgb is None:
if args.verbose:
print(" + Unable to load.")
outRgb = None
else:
outRgb = align.align(imgSize, rgb,
landmarkIndices=landmarkIndices,
skipMulti=not multi)
if outRgb is None and args.verbose:
print(" + Unable to align.")
if outRgb is not None:
if args.verbose:
print(" + Writing aligned file to disk.")
outBgr = cv2.cvtColor(outRgb, cv2.COLOR_RGB2BGR)
cv2.imwrite(imgName, outBgr)
def batchAlign(imgs):
multi = False
start = time.time()
landmarkMap = {
'outerEyesAndNose': openface.AlignDlib.OUTER_EYES_AND_NOSE,
'innerEyesAndBottomLip': openface.AlignDlib.INNER_EYES_AND_BOTTOM_LIP
}
if landmarks not in landmarkMap:
# TODO: Avoid exceptions, find way to silently fail and skip image
raise Exception("Landmarks unrecognized: {}".format(landmarks))
landmarkIndices = landmarkMap[landmarks]
nFallbacks = 0
for imgObject in imgs:
print("=== {} ===".format(imgObject.path))
outDir = os.path.join(alignedDir, imgObject.cls)
mkdirP(outDir)
outputPrefix = os.path.join(outDir, imgObject.name)
imgName = outputPrefix + ".png"
''' print(" + Already found, skipping.")
else:'''
if not os.path.isfile(imgName):
rgb = imgObject.getRGB()
if rgb is None:
print(" + Unable to load.")
outRgb = None
else:
outRgb = align_pred.align(imgSize, rgb,
landmarkIndices=landmarkIndices,
skipMulti=not multi)
if outRgb is None:
print(" + Unable to align.")
if outRgb is not None:
print(" + Writing aligned file to disk.")
outBgr = cv2.cvtColor(outRgb, cv2.COLOR_RGB2BGR)
cv2.imwrite(imgName, outBgr)
def main(args):
mkdirP(args.outputDir)
videos = list(iterVids(args.inputDir))
# TODO - consider shuffling
for videoObject in videos:
print("=== {} ===".format(videoObject.path))
classDir = os.path.join(args.outputDir, videoObject.cls)
mkdirP(classDir)
outDir = os.path.join(classDir, videoObject.name)
mkdirP(outDir)
extract_faces(args, videoObject, outDir)
def prebatchAlign(num_processes, args, multi):
mkdirP(alignedDir)
imgs = list(iterImgs(trainingDir))
np.random.shuffle(imgs)
div_arrays = divideArray(imgs, num_processes)
return div_arrays
def eventDetector(args):
# Make the data folder
if args.verbose:
print ("Creating {} if it does not exist already".format(savedDataDir))
print ("Creating {} if it does not exist already".format(collectBufferDir))
print ("Creating {} if it does not exist already".format(sessionsDir))
mkdirP(savedDataDir)
mkdirP(collectBufferDir)
mkdirP(sessionsDir)
# First load the feed
if len(args.feed) == 0:
if args.verbose:
print("Loading feed from webcam")
cap = cv2.VideoCapture(1)
else:
if args.verbose:
print("Loading feed from {}".format(args.feed))
cap = cv2.VideoCapture(args.feed)
processedFrames = 1
imageCount = 0
idleCount = 0
while (cap.isOpened()):
ret = False
time.sleep(SLEEP_TIME)
while not ret:
ret, frame = cap.read()
processedFrames = processedFrames + 1
# Detect a face
faceDetected = detectFace(frame)
if faceDetected:
idleCount = 0
if (imageCount < NEEDED_IMAGES):
# Save module
savedFacePath = os.path.join(collectBufferDir, "{}.jpg".format(imageCount))
if args.verbose:
print "Saving image to {}".format(savedFacePath)
cv2.imwrite(savedFacePath, frame)
if (imageCount == NEEDED_IMAGES):
newFolderName = time.strftime("%m%d%Y%H%M%S.jpg")
newFolderPath = os.path.join(sessionsDir, newFolderName)
mvP(collectBufferDir, newFolderPath)
if args.verbose:
print "Saved enough images, moving {} to {}".format(collectBufferDir, newFolderPath, newFolderName)
mkdirP(collectBufferDir)
# Move all images in the folder somewhere else
testingThread = threading.Thread(target=testImages, args=(args, newFolderPath))
testingThread.start()
imageCount += 1
else:
# No face detected, count idle
idleCount += 1
if idleCount >= IDLE_WAIT_COUNT:
idleCount = 0
if imageCount < NEEDED_IMAGES:
if args.verbose:
print "Detected no movement for awhile, deleting everything in the folder"
rmdirP(collectBufferDir)
mkdirP(collectBufferDir)
imageCount = 0
cv2.imshow('frame', frame)
k = cv2.waitKey(30) & 0xff
if k == 27:
break
# Update optical flow info
cv2.destroyAllWindows()
cap.release()
def alignMain(args):
util.mkdirP(args.outputDir)
imgs = list(iterImgs(args.inputDir))
# Shuffle so multiple versions can be run at once.
random.shuffle(imgs)
landmarkMap = {
'outerEyesAndNose': AlignDlib.OUTER_EYES_AND_NOSE,
'innerEyesAndBottomLip': AlignDlib.INNER_EYES_AND_BOTTOM_LIP,
'outerEyesAndJaw': AlignDlib.OUTER_EYES_AND_JAW,
'innerEyesAndNose': AlignDlib.INNER_EYES_AND_NOSE
}
if args.landmarks not in landmarkMap:
raise Exception("Landmarks unrecognized: {}".format(args.landmarks))
landmarkIndices = landmarkMap[args.landmarks]
align = AlignDlib(args.dlibFacePredictor)
nFallbacks = 0
for imgObject in imgs:
print("=== {} ===".format(imgObject.path))
outDir = os.path.join(args.outputDir, imgObject.cls)
util.mkdirP(outDir)
outputPrefix = os.path.join(outDir, imgObject.name)
imgName = outputPrefix + ".png"
if os.path.isfile(imgName):
if args.verbose:
print(" + Already found, skipping.")
else:
rgb = imgObject.getRGB()
if rgb is None:
if args.verbose:
print(" + Unable to load.")
outRgb = None
else:
outRgb = align.align(args.size, rgb,
landmarkIndices=landmarkIndices,
skipMulti=args.skipMulti)
if outRgb is None and args.verbose:
print(" + Unable to align.")
if args.fallbackLfw and outRgb is None:
nFallbacks += 1
deepFunneled = "{}/{}.jpg".format(os.path.join(args.fallbackLfw,
imgObject.cls),
imgObject.name)
shutil.copy(deepFunneled, "{}/{}.jpg".format(os.path.join(args.outputDir,
imgObject.cls),
imgObject.name))
if outRgb is not None:
if args.verbose:
print(" + Writing aligned file to disk.")
outBgr = cv2.cvtColor(outRgb, cv2.COLOR_RGB2BGR)
cv2.imwrite(imgName, outBgr)
if args.fallbackLfw:
print('nFallbacks:', nFallbacks)