package/pigo
1
0
Fork 0
mirror of https://github.com/esimov/pigo.git synced 2025-10-06 08:36:54 +08:00
Code Issues Packages Projects Releases Wiki Activity

Implemented the Python part

Browse Source
This commit is contained in:
esimov
2019-10-30 15:05:35 +02:00
parent 0239a7f5a2
commit f61761eb9b
1 changed files with 19 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
examples/talk_detector/talkdet.py
View File

@@ -13,6 +13,8 @@ os.system('rm talkdet.so')
MAX_NDETS = 2024
ARRAY_DIM = 6
MOUTH_AR_THRESH = 0.2
# define class GoPixelSlice to map to:
# C type struct { void *data; GoInt len; GoInt cap; }
class GoPixelSlice(Structure):
@@ -24,16 +26,16 @@ class GoPixelSlice(Structure):
def process_frame(pixs):
dets = np.zeros(ARRAY_DIM * MAX_NDETS, dtype=np.float32)
pixels = cast((c_ubyte * len(pixs))(*pixs), POINTER(c_ubyte))
# call FindFaces
faces = GoPixelSlice(pixels, len(pixs), len(pixs))
pigo.FindFaces.argtypes = [GoPixelSlice]
pigo.FindFaces.restype = c_void_p
# Call the exported FindFaces function from Go.
# Call the exported FindFaces function from Go.
ndets = pigo.FindFaces(faces)
data_pointer = cast(ndets, POINTER((c_longlong * ARRAY_DIM) * MAX_NDETS))
if data_pointer :
buffarr = ((c_longlong * ARRAY_DIM) * MAX_NDETS).from_address(addressof(data_pointer.contents))
res = np.ndarray(buffer=buffarr, dtype=c_longlong, shape=(MAX_NDETS, ARRAY_DIM,))
@@ -42,7 +44,7 @@ def process_frame(pixs):
dets_len = res[0][0]
res = np.delete(res, 0, 0) # delete the first element from the array
# We have to multiply the detection length with the total
# We have to multiply the detection length with the total
# detection points(face, pupils and facial lendmark points), in total 18
dets = list(res.reshape(-1, ARRAY_DIM))[0:dets_len*19]
return dets
@@ -52,7 +54,7 @@ cap = cv2.VideoCapture(0)
cap.set(cv2.CAP_PROP_FRAME_WIDTH, 640)
cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 480)
# Changing the camera resolution introduce a short delay in the camera initialization.
# Changing the camera resolution introduce a short delay in the camera initialization.
# For this reason we should delay the object detection process with a few milliseconds.
time.sleep(0.4)
@@ -71,23 +73,21 @@ while(True):
if dets is not None:
# We know that the detected faces are taking place in the first positions of the multidimensional array.
for det in dets:
if det[3] > 50:
if det[4] == 0: # 0 == face;
for row, col, scale, q, det_type, mouth_ar in dets:
if q > 50:
if det_type == 0: # 0 == face;
if showFaceDet:
cv2.rectangle(frame,
(int(det[1])-int(det[2]/2), int(det[0])-int(det[2]/2)),
(int(det[1])+int(det[2]/2), int(det[0])+int(det[2]/2)),
(0, 0, 255), 2
)
elif det[4] == 1: # 1 == pupil;
cv2.rectangle(frame, (col-scale/2, row-scale/2), (col+scale/2, row+scale/2), (0, 0, 255), 2)
elif det_type == 1: # 1 == pupil;
if showPupil:
cv2.circle(frame, (int(det[1]), int(det[0])), 4, (0, 0, 255), -1, 8, 0)
elif det[4] == 2: # 2 == facial landmark;
cv2.circle(frame, (int(col), int(row)), 4, (0, 0, 255), -1, 8, 0)
elif det_type == 2: # 2 == facial landmark;
if showLandmarkPoints:
cv2.circle(frame, (int(det[1]), int(det[0])), 4, (0, 255, 0), -1, 8, 0)
elif det[4] == 3:
print(det[5])
cv2.circle(frame, (int(col), int(row)), 4, (0, 255, 0), -1, 8, 0)
elif det_type == 3:
if mouth_ar < MOUTH_AR_THRESH: # mouth is open
cv2.putText(frame, "TALKING!", (10, 30),
cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
cv2.imshow('', frame)