def main():
default_model_dir = '../all_models'
default_model = 'mobilenet_ssd_v1_coco_quant_postprocess_edgetpu.tflite'
default_labels = 'coco_labels.txt'
parser = argparse.ArgumentParser()
parser.add_argument('--model',
help='.tflite model path',
default=os.path.join(default_model_dir, default_model))
parser.add_argument('--labels',
help='label file path',
default=os.path.join(default_model_dir,
default_labels))
args = parser.parse_args()
with open(args.labels, 'r') as f:
pairs = (l.strip().split(maxsplit=1) for l in f.readlines())
labels = dict((int(k), v) for k, v in pairs)
engine = edgetpu.detection.engine.DetectionEngine(args.model)
with picamera.PiCamera() as camera:
camera.resolution = (640, 480)
camera.framerate = 30
camera.annotate_text_size = 20
_, width, height, channels = engine.get_input_tensor_shape()
camera.start_preview()
try:
stream = io.BytesIO()
fps = deque(maxlen=20)
fps.append(time.time())
for foo in camera.capture_continuous(
stream,
#format='rgb',
format='jpeg',
use_video_port=True,
resize=(width, height)):
stream.truncate()
stream.seek(0)
# input = np.fromstring(stream.getvalue(),dtype=np.uint8)
camera.capture(stream, format='jpeg')
img = Image.open(stream)
draw = ImageDraw.Draw(img)
output_name = 'images/test_{}.jpeg'.format(time.time())
# Run inference.
ans = engine.DetectWithImage(img,
threshold=0.05,
keep_aspect_ratio=True,
relative_coord=False,
top_k=3)
# Display result.
if ans:
for obj in ans:
print('-----------------------------------------')
if labels:
print(labels[obj.label_id])
print('score = ', obj.score)
box = obj.bounding_box.flatten().tolist()
print('box = ', box)
# Draw a rectangle.
draw.rectangle(box, outline='blue')
draw.text((box[0], box[1]),
'{} {}'.format(labels[obj.label_id],
str(obj.score)), 'red')
if labels[
obj.label_id] == 'person' and obj.score > 0.6:
img.save(output_name)
url = upload_image_datalake(output_name)
short_url = get_bitly_url(url)
push_line_message(short_url)
exit()
img.save(output_name)
if platform.machine() == 'x86_64':
# For gLinux, simply show the image.
img.show()
elif platform.machine() == 'armv7l':
# For Raspberry Pi, you need to install 'feh' to display image.
subprocess.Popen(['feh', output_name])
else:
print('Please check ', output_name)
else:
print('No object detected!')
finally:
camera.stop_preview()
ms_txt = pi3d.String(camera=CAMERA,
is_3d=False,
font=font,
string=ms,
x=0,
y=preview_H / 2 - 30,
z=1.0)
ms_txt.set_shader(txtshader)
last_tm = time.time()
i = 0
with picamera.PiCamera() as camera:
camera.resolution = (preview_W, preview_H)
camera.framerate = max_fps
rgb = PiRGBArray(camera, size=camera.resolution * 3)
_, width, height, channels = engine.get_input_tensor_shape()
camera.start_preview(fullscreen=False,
layer=0,
window=(preview_mid_X, preview_mid_Y, preview_W,
preview_H))
try:
while DISPLAY.loop_running():
stream = io.BytesIO()
camera.capture(stream, use_video_port=True, format='bgr')
stream.truncate()
stream.seek(0)
input = np.frombuffer(stream.getvalue(), dtype=np.uint8)
#stream.close()
start_ms = time.time()
results = engine.DetectWithInputTensor(input, top_k=max_obj)
elapsed_ms = time.time() - start_ms
def main():
parser = argparse.ArgumentParser()
parser.add_argument(
'--model', help='File path of Tflite model.', required=True)
parser.add_argument(
'--dims', help='Model input dimension', required=True)
args = parser.parse_args()
#Set all input params equal to the input dimensions expected by the model
mdl_dims = int(args.dims) #dims must be a factor of 32 for picamera resolution to work
#Set max num of objects you want to detect per frame
max_obj = 10
engine = edgetpu.detection.engine.DetectionEngine(args.model)
pygame.init()
pygame.display.set_caption('Face Detection')
screen = pygame.display.set_mode((mdl_dims, mdl_dims), pygame.DOUBLEBUF|pygame.HWSURFACE)
pygame.font.init()
fnt_sz = 18
myfont = pygame.font.SysFont('Arial', fnt_sz)
camera = picamera.PiCamera()
strm_thread = PiVideoStream().start()
strm_thread.rgbCapture = bytearray(strm_thread.camera.resolution[0] * strm_thread.camera.resolution[1] * 3)
#Set camera resolution equal to model dims
camera.resolution = (mdl_dims, mdl_dims)
#rgb = bytearray(camera.resolution[0] * camera.resolution[1] * 3)
camera.framerate = 30
_, width, height, channels = engine.get_input_tensor_shape()
x1, x2, x3, x4, x5 = 0, 50, 50, 0, 0
y1, y2, y3, y4, y5 = 50, 50, 0, 0, 50
z = 5
last_tm = time.time()
i = 0
exitFlag = True
while(exitFlag):
for event in pygame.event.get():
keys = pygame.key.get_pressed()
if(keys[pygame.K_ESCAPE] == 1):
exitFlag = False
with picamera.array.PiRGBArray(camera, size=(mdl_dims, mdl_dims)) as stream:
#stream = io.BytesIO()
start_ms = time.time()
camera.capture(stream, use_video_port=True, format='rgb')
elapsed_ms = time.time() - start_ms
stream.seek(0)
stream.readinto(stream.rgbCapture)
#stream.truncate() #needed??
img = pygame.image.frombuffer(stream.rgbCapture[0:
(camera.resolution[0] * camera.resolution[1] * 3)],
camera.resolution, 'RGB')
input = np.frombuffer(stream.getvalue(), dtype=np.uint8)
#Inference
results = engine.DetectWithInputTensor(input, top_k=max_obj)
stream.close()
if img:
screen.blit(img, (0,0))
if results:
num_obj = 0
for obj in results:
num_obj = num_obj + 1
for obj in results:
bbox = obj.bounding_box.flatten().tolist()
score = round(obj.score,2)
x1 = round(bbox[0] * mdl_dims)
y1 = round(bbox[1] * mdl_dims)
x2 = round(bbox[2] * mdl_dims)
y2 = round(bbox[3] * mdl_dims)
rect_width = x2 - x1
rect_height = y2 - y1
class_score = "%.2f" % (score)
ms = "(%d) %s%.2fms" % (num_obj, "faces detected in ", elapsed_ms*1000)
fnt_class_score = myfont.render(class_score, True, (0,0,255))
fnt_class_score_width = fnt_class_score.get_rect().width
screen.blit(fnt_class_score,(x1, y1-fnt_sz))
fnt_ms = myfont.render(ms, True, (255,255,255))
fnt_ms_width = fnt_ms.get_rect().width
screen.blit(fnt_ms,((mdl_dims / 2) - (fnt_ms_width / 2), 0))
bbox_rect = pygame.draw.rect(screen, (0,0,255), (x1, y1, rect_width, rect_height), 2)
#pygame.display.update(bbox_rect)
else:
elapsed_ms = time.time() - start_ms
ms = "%s %.2fms" % ("No faces detected in", elapsed_ms*1000)
fnt_ms = myfont.render(ms, True, (255,0,0))
fnt_ms_width = fnt_ms.get_rect().width
screen.blit(fnt_ms,((mdl_dims / 2) - (fnt_ms_width / 2), 0))
pygame.display.update()
pygame.display.quit()