def classify_frame(img, inputQueue, outputQueue):
engine = edgetpu.detection.engine.DetectionEngine(\
'unquant_tpu_tflite/landing_tflite_graph.tflite')
# keep looping
while True:
# check to see if there is a frame in our input queue
if not inputQueue.empty():
# grab the frame from the input queue
img = inputQueue.get()
results = engine.DetectWithImage(img, threshold=0.4,\
keep_aspect_ratio=True, relative_coord=False, top_k=10)
data_out = []
if results:
for obj in results:
inference = []
box = obj.bounding_box.flatten().tolist()
xmin = int(box[0])
ymin = int(box[1])
xmax = int(box[2])
ymax = int(box[3])
inference.extend(
(obj.label_id, obj.score, xmin, ymin, xmax, ymax))
data_out.append(inference)
outputQueue.put(data_out)
def predict(image, class_to_detect, confidence):
image_height, image_width = image.shape[:2]
image = Image.fromarray(image)
results = engine.DetectWithImage(image,
threshold=0.4,
keep_aspect_ratio=True,
relative_coord=False,
top_k=10)
predictions = []
if results:
for obj in results:
box = obj.bounding_box.flatten().tolist()
xmin = int(box[0])
ymin = int(box[1])
xmax = int(box[2])
ymax = int(box[3])
if obj.score > confidence and obj.label_id == class_to_detect.value:
predictions.append(
Prediction(class_id=obj.label_id,
img_w=image_width,
img_h=image_height,
bounding_box=Box(xmin, ymin, xmax - xmin,
ymax - ymin)))
return predictions
def main():
os.chdir('/home/pi/DeepPiCar/models/object_detection')
parser = argparse.ArgumentParser()
parser.add_argument('--model',
help='File path of Tflite model.',
required=False)
parser.add_argument('--label',
help='File path of label file.',
required=False)
args = parser.parse_args()
#args.model = 'test_data/mobilenet_ssd_v2_coco_quant_postprocess.tflite'
args.model = 'data/model_result/road_signs_quantized.tflite'
args.label = 'data/model_result/road_sign_labels.txt'
with open(args.label, 'r') as f:
pairs = (l.strip().split(maxsplit=1) for l in f.readlines())
labels = dict((int(k), v) for k, v in pairs)
# initialize open cv
IM_WIDTH = 640
IM_HEIGHT = 480
camera = cv2.VideoCapture(0)
ret = camera.set(3, IM_WIDTH)
ret = camera.set(4, IM_HEIGHT)
font = cv2.FONT_HERSHEY_SIMPLEX
bottomLeftCornerOfText = (10, IM_HEIGHT - 10)
fontScale = 1
fontColor = (255, 255, 255) # white
boxColor = (0, 0, 255) # RED?
boxLineWidth = 1
lineType = 2
annotate_text = ""
annotate_text_time = time.time()
time_to_show_prediction = 1.0 # ms
min_confidence = 0.20
# initial classification engine
engine = edgetpu.detection.engine.DetectionEngine(args.model)
elapsed_ms = 0
fourcc = cv2.VideoWriter_fourcc(*'XVID')
out = cv2.VideoWriter('output.avi', fourcc, 20.0, (IM_WIDTH, IM_HEIGHT))
try:
while camera.isOpened():
try:
start_ms = time.time()
ret, frame = camera.read() # grab a frame from camera
if ret == False:
print('can NOT read from camera')
break
frame_expanded = np.expand_dims(frame, axis=0)
ret, img = camera.read()
input = cv2.cvtColor(
frame, cv2.COLOR_BGR2RGB) # convert to RGB color space
img_pil = Image.fromarray(input)
#input = cv2.resize(input, (width,height))
start_tf_ms = time.time()
results = engine.DetectWithImage(img_pil,
threshold=min_confidence,
keep_aspect_ratio=True,
relative_coord=False,
top_k=5)
end_tf_ms = time.time()
elapsed_tf_ms = end_tf_ms - start_ms
if results:
for obj in results:
print("%s, %.0f%% %s %.2fms" %
(labels[obj.label_id], obj.score * 100,
obj.bounding_box, elapsed_tf_ms * 1000))
box = obj.bounding_box
coord_top_left = (int(box[0][0]), int(box[0][1]))
coord_bottom_right = (int(box[1][0]), int(box[1][1]))
cv2.rectangle(img, coord_top_left, coord_bottom_right,
boxColor, boxLineWidth)
annotate_text = "%s, %.0f%%" % (labels[obj.label_id],
obj.score * 100)
coord_top_left = (coord_top_left[0],
coord_top_left[1] + 15)
cv2.putText(img, annotate_text, coord_top_left, font,
fontScale, boxColor, lineType)
print('------')
else:
print('No object detected')
# Print Frame rate info
elapsed_ms = time.time() - start_ms
annotate_text = "%.2f FPS, %.2fms total, %.2fms in tf " % (
1.0 / elapsed_ms, elapsed_ms * 1000, elapsed_tf_ms * 1000)
print('%s: %s' % (datetime.datetime.now(), annotate_text))
cv2.putText(img, annotate_text, bottomLeftCornerOfText, font,
fontScale, fontColor, lineType)
out.write(img)
cv2.imshow('Detected Objects', img)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
except:
# catch it and don't exit the while loop
print('In except')
traceback.print_exc()
finally:
print('In Finally')
camera.release()
out.release()
cv2.destroyAllWindows()
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()
import time
import re
import numpy as np
import os
import datetime
import time
import edgetpu.detection.engine
from PIL import Image
model = "road_signs_quantized_v2_edgetpu.tflite"
label = 'road_sign_labels.txt'
image_path = "data/test/propios/IMG_20190715_212222.jpg"
with open(label, '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(model)
start = time.time()
image = Image.open(image_path)
results = engine.DetectWithImage(image, threshold=0.20, top_k=6)
end = time.time()
for result in results:
print(labels[result.label_id], result.score * 100)
#end = time.time()
print(end - start)
def main() -> None:
# Store labels for matching with inference results
labels = read_label_file(ARGS.labels) if ARGS.labels else None
# Specify font for labels
# font = PIL.ImageFont.truetype("/usr/share/fonts/truetype/piboto/Piboto-Regular.ttf", 20)
font = PIL.ImageFont.truetype(
"/usr/share/fonts/truetype/roboto/hinted/Roboto-Regular.ttf", 30)
# font = None
engine = edgetpu.detection.engine.DetectionEngine(ARGS.model)
width = 640
height = 480
out = cv2.VideoWriter('output.avi',
cv2.VideoWriter_fourcc(*'MJPG'),
30,
(width, height))
# Use imutils to count Frames Per Second (FPS)
fps = FPS().start()
live_view = LiveView(ARGS.ip, ARGS.port)
while True:
try:
try:
image = PIL.Image.open(io.BytesIO(live_view.image()))
except OSError:
print('could not identify image file')
continue
# Perform inference and note time taken
startMs = time.time()
try:
inferenceResults = engine.DetectWithImage(image,
threshold=ARGS.confidence,
keep_aspect_ratio=True,
relative_coord=False,
top_k=ARGS.maxobjects)
elapsedMs = time.time() - startMs
annotate(image, inferenceResults, elapsedMs, labels, font)
except OSError as e:
print(e)
pass
cv2_image = cv2.cvtColor(numpy.asarray(image), cv2.COLOR_RGB2BGR)
out.write(cv2_image)
if 'DISPLAY' in os.environ:
cv2.imshow('NCS Improved live inference', cv2_image)
# Display the frame for 5ms, and close the window so that the next
# frame can be displayed. Close the window if 'q' or 'Q' is pressed.
if cv2.waitKey(5) & 0xFF == ord('q'):
fps.stop()
break
fps.update()
# Allows graceful exit using ctrl-c (handy for headless mode).
except KeyboardInterrupt:
fps.stop()
break
print("Elapsed time: " + str(fps.elapsed()))
print("Approx FPS: :" + str(fps.fps()))
cv2.destroyAllWindows()