def showScreenAndDectect(capture):
print("Face dececting...")
cnt = 5
while cnt:
flag, frame = capture.read()
faceCoordinates = fdu.getFaceCoordinates(frame)
refreshFrame(frame, faceCoordinates)
if faceCoordinates is not None:
cnt -= 1
face_img = fdu.preprocess(frame,
faceCoordinates,
face_shape=FACE_SHAPE)
#cv2.imshow(windowsName, face_img)
input_img = np.expand_dims(face_img, axis=0)
input_img = np.expand_dims(input_img, axis=0)
result = model.predict(input_img)[0]
if cnt == 4:
tot_result = result
else:
tot_result += result
index = np.argmax(result)
print('Frame', 5 - cnt, ':', emo[index], 'prob:', max(result))
#index = np.argmax(result)
#print (emo[index], 'prob:', max(result))
# print(face_img.shape)
# emotion = class_label[result_index]
# print(emotion)
index = np.argmax(tot_result)
print('Final decision:', emo[index], 'prob:', max(tot_result))
return emo[index]
def showScreenAndDectect(capture):
while (True):
flag, frame = capture.read()
faceCoordinates = fdu.getFaceCoordinates(frame)
refreshFrame(frame, faceCoordinates)
if faceCoordinates is not None:
face_img = fdu.preprocess(frame, faceCoordinates, face_shape=FACE_SHAPE)
#cv2.imshow(windowsName, face_img)
input_img = np.expand_dims(face_img, axis=0)
input_img = np.expand_dims(input_img, axis=0)
result = model.predict(input_img)[0]
index = np.argmax(result)
jsonObj = {
"Angry" : result[0],
"Fear" : result[1],
"Happy" : result[2];
"Sad" : result[3],
"Suprise" : result[4],
"Neutral" : result[5]
}
r = requests.get(url = API_ENDPOINT, data = jsonObj)
pause(1)
print (emo[index], 'prob:', max(result))
def showScreenAndDectect(capture):
while (True):
flag, frame = capture.read()
faceCoordinates = fdu.getFaceCoordinates(frame)
refreshFrame(frame, faceCoordinates)
if faceCoordinates is not None:
face_img = fdu.preprocess(frame, faceCoordinates, face_shape=FACE_SHAPE)
#cv2.imshow(windowsName, face_img)
input_img = np.expand_dims(face_img, axis=0)
input_img = np.expand_dims(input_img, axis=0)
result = model.predict(input_img)[0]
index = np.argmax(result)
print (emo[index], 'prob:', max(result))
def getEmotion(self):
i = 0
capture = self.getCapture()
seq = []
while (i < 5):
flag, frame = capture.read()
faceCoordinates = fdu.getFaceCoordinates(frame)
if faceCoordinates is not None:
face_img = fdu.preprocess(frame,
faceCoordinates,
face_shape=FACE_SHAPE)
input_img = np.expand_dims(face_img, axis=0)
input_img = np.expand_dims(input_img, axis=0)
result = model.predict(input_img)[0]
index = np.argmax(result)
print(emo[index], 'prob:', max(result))
seq.append(emo[index])
i += 1
return seq[2]
def showScreenAndDectect(capture):
while (True):
'''
Capture frame-by-frame
read() returns two parameters:
1.The actual video frame read (one frame on each loop)
2.A return code (The return code tells us if we have run out of frames, which will happen if we are reading from a file. This doesn’t matter when reading from the webcam, since we can record forever, so we will ignore it.)
'''
flag, frame = capture.read()
faceCoordinates = fdu.getFaceCoordinates(frame)
# refreshFrame(frame, faceCoordinates)
# Another solution, get multiple faces to show here.
'''
flag, frame = capture.read()
faceCoordinates = fdu.getFaceCoordinates2(frame)
refreshFrame2(frame, faceCoordinates)
'''
if faceCoordinates is not None:
face_img = fdu.preprocess(frame,
faceCoordinates,
face_shape=FACE_SHAPE)
#cv2.imshow(windowsName, face_img)
input_img = np.expand_dims(face_img, axis=0)
input_img = np.expand_dims(input_img, axis=0)
result = model.predict(input_img)[0]
index = np.argmax(result)
print(emo[index], 'probability is: ', max(result))
refreshFrame(frame, faceCoordinates, emo[index])
if cv2.waitKey(1) & 0xFF == ord('q'):
break
# When everything is done, release the capture
capture.release()
cv2.destroyAllWindows()
def sendCertifyResult(self, dataURL, identity):
head = "data:image/jpeg;base64,"
assert (dataURL.startswith(head))
imgdata = base64.b64decode(dataURL[len(head):])
imgF = StringIO.StringIO()
imgF.write(imgdata)
imgF.seek(0)
img = Image.open(imgF)
# preprocess frame
buf = np.fliplr(np.asarray(img))
rgbFrame = np.zeros((300, 400, 3), dtype=np.uint8)
rgbFrame[:, :, 0] = buf[:, :, 0]
rgbFrame[:, :, 1] = buf[:, :, 1]
rgbFrame[:, :, 2] = buf[:, :, 2]
'''
rgbFrame[:, :, 0] = buf[:, :, 2]
rgbFrame[:, :, 1] = buf[:, :, 1]
rgbFrame[:, :, 2] = buf[:, :, 0]
'''
#rgbFrame = cv2.resize(rgbFrame, (320, 240))
#print (rgbFrame)
#alignedFace = cv2.cvtColor(rgbFrame, cv2.COLOR_RGB2GRAY)
fcoord = fdu.getFaceCoordinates(rgbFrame)
print(fcoord)
if fcoord is None:
return
alignedFace = fdu.preprocess(rgbFrame, fcoord, (64, 64))
cv2.imwrite("test.png", alignedFace)
alignedFace = np.expand_dims(alignedFace, axis=-1)
print(alignedFace.shape)
'''
bb = align.getLargestFaceBoundingBox(rgbFrame)
bbs = [bb] if bb is not None else []
for bb in bbs:
# print(len(bbs))
landmarks = align.findLandmarks(rgbFrame, bb)
alignedFace = align.align(args.imgDim, rgbFrame, bb,
landmarks=landmarks,
landmarkIndices=openface.AlignDlib.OUTER_EYES_AND_NOSE)
alignedFace = cv2.cvtColor(alignedFace, cv2.COLOR_RGB2GRAY)
alignedFace = cv2.resize(alignedFace, (48, 48))
cv2.imwrite("test.png", alignedFace)
alignedFace = np.expand_dims(alignedFace, axis=0)
#print (alignedFace.shape)
'''
if alignedFace is None:
return
self.certifying_buf.append(alignedFace / 255.0)
if (len(self.certifying_buf) >= 15):
frame_to_be_tested = np.asarray(self.certifying_buf)
result = self.emo_model.predict(frame_to_be_tested)
print(result)
print(result.sum(0))
emotion_type = np.argmax(result.sum(0))
if emotion_type == 3 and identity != -1:
# send success signal
print("successful")
msg = {"type": "CERTIFIED_SUCCESS"}
elif identity == -1:
msg = {"type": "CERTIFIED_FAIL", "val": 1}
else:
msg = {"type": "CERTIFIED_FAIL", "val": 0}
self.sendMessage(json.dumps(msg))
# empty buffer
self.certifying_buf = []
# reset certifying signal
self.certifying = False
