def main():
"""Run csi video stream and object detector."""
obj_detect = edgeiq.ObjectDetection(
"alwaysai/ssd_mobilenet_v1_coco_2018_01_28")
obj_detect.load(engine=edgeiq.Engine.DNN_CUDA)
print("Loaded model:\n{}\n".format(obj_detect.model_id))
print("Engine: {}".format(obj_detect.engine))
print("Accelerator: {}\n".format(obj_detect.accelerator))
print("Labels:\n{}\n".format(obj_detect.labels))
try:
with enhanced_csi.JetsonVideoStream(cam=0,
rotation=enhanced_csi.
FrameRotation.ROTATE_180,
camera_mode=enhanced_csi.
JetsonCameraMode.
IMX477_4032x3040_30_0,
display_width=640,
display_height=480) as video_stream,\
edgeiq.Streamer() as streamer:
time.sleep(2.0)
video_stream.start_counting_fps()
# loop detection
while True:
frame = enhanced_csi.read_camera(video_stream, True)
results = obj_detect.detect_objects(frame, confidence_level=.4)
frame = edgeiq.markup_image(frame,
results.predictions,
colors=obj_detect.colors)
# Generate text to display on streamer
text = ["Model: {}".format(obj_detect.model_id)]
text.append("Inference time: {:1.3f} s".format(
results.duration))
text.append("Objects:")
for prediction in results.predictions:
text.append("{}: {:2.2f}%".format(
prediction.label, prediction.confidence * 100))
video_stream.frames_displayed += 1
streamer.send_data(frame, text)
if streamer.check_exit():
break
video_stream.release_fps_stats()
finally:
print("Program Ending")
def main():
facial_detector = edgeiq.ObjectDetection(
"alwaysai/res10_300x300_ssd_iter_140000")
facial_detector.load(engine=edgeiq.Engine.DNN)
print("Engine: {}".format(facial_detector.engine))
print("Accelerator: {}\n".format(facial_detector.accelerator))
print("Model:\n{}\n".format(facial_detector.model_id))
fps = edgeiq.FPS()
try:
with edgeiq.WebcamVideoStream(cam=0) as webcam, \
edgeiq.Streamer() as streamer:
# Allow webcam to warm up
time.sleep(2.0)
fps.start()
# loop detection
while True:
frame = webcam.read()
# detect human faces
results = facial_detector.detect_objects(frame,
confidence_level=.5)
frame = edgeiq.markup_image(frame,
results.predictions,
colors=facial_detector.colors)
# Generate text to display on streamer
text = ["Model: {}".format(facial_detector.model_id)]
text.append("Inference time: {:1.3f} s".format(
results.duration))
text.append("Faces:")
for prediction in results.predictions:
text.append("{}: {:2.2f}%".format(
prediction.label, prediction.confidence * 100))
streamer.send_data(frame, text)
fps.update()
if streamer.check_exit():
break
finally:
# stop fps counter and display information
fps.stop()
print("[INFO] elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("[INFO] approx. FPS: {:.2f}".format(fps.compute_fps()))
print("Program Ending")
def main():
obj_detect = edgeiq.ObjectDetection(
"alwaysai/ssd_mobilenet_v2_coco_2018_03_29")
obj_detect.load(engine=edgeiq.Engine.DNN)
print("Loaded model:\n{}\n".format(obj_detect.model_id))
print("Engine: {}".format(obj_detect.engine))
print("Accelerator: {}\n".format(obj_detect.accelerator))
print("Labels:\n{}\n".format(obj_detect.labels))
try:
with edgeiq.RealSense() as video_stream, \
edgeiq.Streamer() as streamer:
print("starting RealSense camera")
time.sleep(2.0)
# loop detection
while True:
distances = []
depth_image, color_image = video_stream.read()
roi = video_stream.roi(depth_image, color_image, min=None, max=0.9)
# frame = edgeiq.resize(color_image, width=416)
results = obj_detect.detect_objects(roi, confidence_level=.6)
roi = edgeiq.markup_image(
roi, results.predictions, colors=obj_detect.colors)
for prediction in results.predictions:
distances.append(video_stream.compute_object_distance(prediction.box,depth_image))
# Generate text to display on streamer
text = ["Model: {}".format(obj_detect.model_id)]
text.append(
"Inference time: {:1.3f} s".format(results.duration))
text.append("Objects:")
for i, prediction in enumerate(results.predictions):
text.append("{}: {:2.1f}% Distance = {:2.2f}m".format(
prediction.label, prediction.confidence * 100, distances[i]))
streamer.send_data(roi, text)
if streamer.check_exit():
break
finally:
print("Program Ending")
def main():
obj_detect = edgeiq.ObjectDetection(
"alwaysai/ssd_mobilenet_v2_coco_2018_03_29")
obj_detect.load(engine=edgeiq.Engine.DNN)
print("Loaded model:\n{}\n".format(obj_detect.model_id))
print("Engine: {}".format(obj_detect.engine))
print("Accelerator: {}\n".format(obj_detect.accelerator))
print("Labels:\n{}\n".format(obj_detect.labels))
fps = edgeiq.FPS()
try:
with edgeiq.WebcamVideoStream(cam=0) as video_stream, \
edgeiq.Streamer() as streamer:
# Allow Webcam to warm up
time.sleep(2.0)
fps.start()
# loop detection
while True:
frame = video_stream.read()
results = obj_detect.detect_objects(frame, confidence_level=.5)
frame = edgeiq.markup_image(frame,
results.predictions,
colors=obj_detect.colors)
# Generate text to display on streamer
text = ["Model: {}".format(obj_detect.model_id)]
text.append("Inference time: {:1.3f} s".format(
results.duration))
text.append("Objects:")
for prediction in results.predictions:
text.append("{}: {:2.2f}%".format(
prediction.label, prediction.confidence * 100))
streamer.send_data(frame, text)
fps.update()
if streamer.check_exit():
break
finally:
fps.stop()
print("elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("approx. FPS: {:.2f}".format(fps.compute_fps()))
print("Program Ending")
def main():
text = "Facial Landmarks with Dlib"
fps = edgeiq.FPS()
shape_predictor = "shape_predictor_68_face_landmarks.dat"
dlib_flm = facial_landmarks.Dlib_FLM(shape_predictor)
try:
with edgeiq.WebcamVideoStream(cam=0) as webcam, \
edgeiq.Streamer() as streamer:
# Allow webcam to warm up
time.sleep(2.0)
fps.start()
# loop detection
while True:
frame = webcam.read()
resized_frame, gray_resized_frame = dlib_flm.image_preprocessor(frame)
facial_coordinates, rectangles =dlib_flm.detect_faces_shapes(gray_resized_frame)
# Loop to markup resized_frame
for(i, rectangle) in enumerate(rectangles):
(x, y, w, h) = dlib_flm.dlib_rectangle_to_cv_bondingbox(rectangle)
cv2.rectangle(resized_frame, (x, y), (x + w, y + h), (0, 255, 0), 2)
cv2.putText(resized_frame, "Face #{}".format(i + 1), (x - 10, y - 10),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)
for facial_coordinate in facial_coordinates:
for (x,y) in facial_coordinate:
cv2.circle(resized_frame, (x, y), 1, (255, 0, 0), -1)
streamer.send_data(resized_frame, text)
fps.update()
if streamer.check_exit():
break
finally:
# stop fps counter and display information
fps.stop()
print("[INFO] elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("[INFO] approx. FPS: {:.2f}".format(fps.compute_fps()))
print("Program Ending")
def main():
obj_det = edgeiq.ObjectDetection("tester2204/CE-Recog")
if edgeiq.is_jetson():
obj_det.load(engine=edgeiq.Engine.DNN_CUDA)
print("Nvidia Jetson Detected\n")
else:
obj_det.load(engine=edgeiq.Engine.DNN)
print("Device is not a Nvidia Jetson Board\n")
print("Initializing Application...\n")
print("Model:\n{}\n".format(obj_det.model_id))
print("Engine:\n{}\n".format(obj_det.engine))
print("Labels:\n{}\n".format(obj_det.labels))
#imgURL = "https://specials-images.forbesimg.com/imageserve/5e88b867e2bb040006427704/0x0.jpg"
#urllib.request.urlretrieve(imgURL, "this.jpg") #Change based on OS and User
#image = "Images/this.jpg"
image_lists = sorted(list(edgeiq.list_images("Images/")))
with edgeiq.Streamer(queue_depth=len(image_lists),
inter_msg_time=7) as streamer:
i = 0
while i < 3:
for image_list in image_lists:
show_image = cv2.imread(image_list)
image = show_image.copy()
results = obj_det.detect_objects(image, confidence_level=.5)
image = edgeiq.markup_image(image,
results.predictions,
colors=obj_det.colors)
shown = ["Model: {}".format(obj_det.model_id)]
shown.append("Inference time: {:1.3f} s".format(
results.duration))
shown.append("Objects:")
for prediction in results.predictions:
shown.append("{}: {:2.2f}%".format(
prediction.label, prediction.confidence * 100))
streamer.send_data(image, shown)
streamer.wait()
i = i + 1
#if streamer.check_exit():
print("That's it folks!")
print("Thanks for using Ben's Object Recognition Model & Software")
print("Sponsored by: Darien's Face")
def main():
cameras = []
# This is the list of camera indices to use. If you increase the length of this list,
# you'll also need to update the `concatenate` step for displaying the frames.
camera_idxs = [0, 1]
for i in camera_idxs:
cameras.append(
CameraThread(i, edgeiq.Engine.DNN, "alwaysai/mobilenet_ssd"))
for c in cameras:
c.start()
try:
with edgeiq.Streamer() as streamer:
while True:
results = []
for c in cameras:
results.append(c.get_results())
# Generate text to display on streamer
text = []
for r in results:
if r is not None:
text.append("Camera {}:".format(r["idx"]))
text.append("Model: {}".format(r["model_id"]))
text.append("Inference time: {:1.3f} s".format(
r["results"].duration))
text.append("Objects:")
for prediction in r["results"].predictions:
text.append("{}: {:2.2f}%".format(
prediction.label, prediction.confidence * 100))
# Join the incoming frames vertically into a single image to be shown on
# the Streamer
frame = np.concatenate(
(results[0]["frame"], results[1]["frame"]), axis=0)
streamer.send_data(frame, text)
if streamer.check_exit():
break
finally:
for c in cameras:
c.stop()
c.join()
print("Program Ending")
def main():
pose_estimator = edgeiq.PoseEstimation("alwaysai/human-pose")
pose_estimator.load(engine=edgeiq.Engine.DNN_OPENVINO)
print("Loaded model:\n{}\n".format(pose_estimator.model_id))
print("Engine: {}".format(pose_estimator.engine))
print("Accelerator: {}\n".format(pose_estimator.accelerator))
fps = edgeiq.FPS()
try:
with edgeiq.WebcamVideoStream(cam=0) as video_stream, \
edgeiq.Streamer() as streamer:
# Allow Webcam to warm up
time.sleep(2.0)
fps.start()
# loop detection
while True:
frame = video_stream.read()
results = pose_estimator.estimate(frame)
pdict = dict()
myList = [None] * 15
# Generate text to display on streamer
text = ["Model: {}".format(pose_estimator.model_id)]
text.append(
"Inference time: {:1.3f} s".format(results.duration) +
"\nFPS: {:.2f}".format(fps.compute_fps()))
for ind, pose in enumerate(results.poses):
pdict["Person {}".format(ind)] = pose.key_points
df = pd.DataFrame(data=pdict,
index=pose.key_points,
columns=pdict)
print(df)
streamer.send_data(results.draw_poses(frame), text)
fps.update()
if streamer.check_exit():
break
finally:
fps.stop()
print("elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("approx. FPS: {:.2f}".format(fps.compute_fps()))
print("Program Ending")
def main():
semantic_segmentation = edgeiq.SemanticSegmentation("alwaysai/enet")
semantic_segmentation.load(engine=edgeiq.Engine.DNN_CUDA)
print("Loaded model:\n{}\n".format(semantic_segmentation.model_id))
print("Engine: {}".format(semantic_segmentation.engine))
print("Accelerator: {}\n".format(semantic_segmentation.accelerator))
print("Labels:\n{}\n".format(semantic_segmentation.labels))
fps = edgeiq.FPS()
try:
with edgeiq.FileVideoStream('toronto.mp4', play_realtime=True) as video_stream, \
edgeiq.Streamer() as streamer: # play_realtime simulates video feed from a camera
# Allow Webcam to warm up
time.sleep(2.0)
fps.start()
# loop detection
while True:
frame = video_stream.read()
results = semantic_segmentation.segment_image(frame)
# Generate text to display on streamer
text = ["Model: {}".format(semantic_segmentation.model_id)]
text.append("Inference time: {:1.3f} s".format(
results.duration))
text.append("Legend:")
text.append(semantic_segmentation.build_legend())
mask = semantic_segmentation.build_image_mask(
results.class_map)
blended = edgeiq.blend_images(frame, mask, alpha=0.5)
streamer.send_data(blended, text)
fps.update()
if streamer.check_exit():
break
finally:
fps.stop()
print("elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("approx. FPS: {:.2f}".format(fps.compute_fps()))
print("Program Ending")
def main():
classifier = edgeiq.Classification("alwaysai/googlenet")
classifier.load(engine=edgeiq.Engine.DNN)
print("Engine: {}".format(classifier.engine))
print("Accelerator: {}\n".format(classifier.accelerator))
print("Model:\n{}\n".format(classifier.model_id))
print("Labels:\n{}\n".format(classifier.labels))
image_paths = sorted(list(edgeiq.list_images("images/")))
print("Images:\n{}\n".format(image_paths))
with edgeiq.Streamer(
queue_depth=len(image_paths), inter_msg_time=3) as streamer:
black_img= cv2.imread('black.jpg')
for image_path in image_paths:
image_display = cv2.imread(image_path)
image = image_display.copy()
results = classifier.classify_image(image)
# Generate text to display on streamer
text = ["Model: {}".format(classifier.model_id)]
text.append("Inference time: {:1.3f} s".format(results.duration))
if results.predictions:
image_text = "Label: {}, {:.2f}".format(
results.predictions[0].label,
results.predictions[0].confidence)
cv2.putText(
image_display, image_text, (5, 25),
cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
for idx, prediction in enumerate(results.predictions[:5]):
text.append("{}. label: {}, confidence: {:.5}".format(
idx + 1, prediction.label, prediction.confidence))
if prediction.label == "Not Safe For Work":
resized_black_image = edge_tools.resize(black_img, image.shape[1], image.shape[0], keep_scale=False)
image_display = edge_tools.blend_images(resized_black_image, image, 0.1)
else:
text.append("No classification for this image.")
streamer.send_data(image_display, text)
streamer.wait()
print("Program Ending")
def main():
fps = edgeiq.FPS()
try:
with edgeiq.WebcamVideoStream(cam=1) as video_stream, \
edgeiq.Streamer() as streamer:
# Allow Webcam to warm up
time.sleep(2.0)
fps.start()
# loop detection
while True:
frame = video_stream.read()
# HSV
frame_hsv = cv2.cvtColor(frame, cv2.COLOR_RGB2HSV)
frame_value = frame_hsv[:, :, 2]
# bilateral filter - edge-preserving image smoothing method
blurredBrightness = cv2.bilateralFilter(
frame_value, 9, 150, 150)
# Canny edge detector
thresh = 50
edges = cv2.Canny(blurredBrightness,
thresh,
thresh * 2,
L2gradient=True)
# Generate text to display on streamer
text = "Thermal Edge Detector"
streamer.send_data(edges, text)
fps.update()
if streamer.check_exit():
break
finally:
fps.stop()
print("elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("approx. FPS: {:.2f}".format(fps.compute_fps()))
print("Program Ending")
def main():
"""Run object detector and centroid tracker."""
tracker = edgeiq.CorrelationTracker(
max_objects=5, enter_cb=face_enters, exit_cb=face_exits)
fps = edgeiq.FPS()
try:
with edgeiq.oak.Oak('alwaysai/face_detection_0200_oak',
sensor=edgeiq.Sensor.res_1080,
video_mode=edgeiq.VideoMode.preview) as oak_camera, \
edgeiq.Streamer() as streamer:
# Allow Webcam to warm up
time.sleep(2.0)
fps.start()
# loop detection and tracking
while True:
frame = oak_camera.get_frame()
results = oak_camera.get_model_result(confidence_level=.6)
if results:
fps.update()
text = ["Faces Detected:"]
objects = tracker.update(results.predictions, frame)
# Update the label to reflect the object ID
predictions = []
for (object_id, prediction) in objects.items():
prediction.label = "face {}".format(object_id)
text.append("{}".format(prediction.label))
predictions.append(prediction)
text.append(("approx. FPS: {:.2f}".
format(fps.compute_fps())))
frame = edgeiq.markup_image(frame, predictions)
streamer.send_data(frame, text)
if streamer.check_exit():
break
finally:
fps.stop()
print("elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("Program Ending")
def main():
"""Run object detector."""
fps = edgeiq.FPS()
try:
with edgeiq.Oak('alwaysai/ssd_v2_coco_oak') as oak_camera,\
edgeiq.Streamer() as streamer:
# Allow Oak camera to warm up
time.sleep(2.0)
fps.start()
# loop detection
while True:
frame = oak_camera.get_frame()
results = oak_camera.get_model_result(confidence_level=.8)
if results:
fps.update()
text = ["Oak Camera Detections:"]
text.append("approx. FPS: {:.2f}".format(
fps.compute_fps()))
text.append("Objects:")
for prediction in results.predictions:
center = tuple(
int(round(val)) for val in prediction.box.center)
b, g, r = frame[center[1], center[0]]
cname = getColorName(r, g, b)
text.append("{}: {:2.2f}% color = {}".format(
prediction.label, prediction.confidence * 100,
cname))
# Mark up image for display
frame = edgeiq.markup_image(frame, results.predictions)
streamer.send_data(frame, text)
if streamer.check_exit():
break
finally:
fps.stop()
print("elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("Program Ending")
def main():
fps = edgeiq.FPS()
try:
streamer = edgeiq.Streamer()
streamer.setup()
video_stream = edgeiq.WebcamVideoStream(
cam=0) # replace with FileVideoStream if need be
# Allow application to warm up
video_stream.start()
time.sleep(2.0)
fps.start()
text = [""]
# initialize Vaccine Trakcer
vaccine_tracker = VaccineTracker()
# loop detection
while True:
frame = video_stream.read()
vaccine_tracker.update(frame)
# draw the vaccination box in the frame
frame = edgeiq.markup_image(frame, [
edgeiq.ObjectDetectionPrediction(label="vaccination",
index=0,
box=vaccine_tracker.box,
confidence=100.00)
])
streamer.send_data(frame, text)
fps.update()
if streamer.check_exit():
break
finally:
fps.stop()
streamer.close()
video_stream.stop()
print("elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("approx. FPS: {:.2f}".format(fps.compute_fps()))
print("Program Ending")
def pose_estimation():
"""
Oak.get_model_result can return different results based on the purpose of the model running on the camera.
This function shows how to work with pose estimation models.
"""
fps = edgeiq.FPS()
with edgeiq.Oak('alwaysai/human_pose_oak') as camera, edgeiq.Streamer(
) as streamer:
fps.start()
while True:
text = ['FPS: {:2.2f}'.format(fps.compute_fps())]
frame = camera.get_frame()
result = camera.get_model_result()
# Check for inferencing results. Oak.get_model_result is a non-blocking call and will return None when new data is not available.
if result:
frame = result.draw_poses(frame)
text.append("Poses:")
for ind, pose in enumerate(result.poses):
text.append("Person {}".format(ind))
text.append('-' * 10)
text.append("Key Points:")
for key_point in pose.key_points:
text.append(str(key_point))
streamer.send_data(frame, text)
if streamer.check_exit():
break
fps.update()
print('fps = {}'.format(fps.compute_fps()))
def depth_stream():
"""
This function shows how to retrieve the depth stream from the camera.
"""
fps = edgeiq.FPS()
with Oak('alwaysai/mobilenet_ssd_oak',
capture_depth=True) as camera, edgeiq.Streamer() as streamer:
fps.start()
while True:
text = ['FPS: {:2.2f}'.format(fps.compute_fps())]
depth = camera.get_depth()
if depth is not None:
streamer.send_data(depth, text)
fps.update()
if streamer.check_exit():
break
print('fps = {}'.format(fps.compute_fps()))
def main(camera, use_streamer, server_addr, stream_fps):
fps = edgeiq.FPS()
try:
streamer = None
if use_streamer:
streamer = edgeiq.Streamer().setup()
else:
streamer = CVClient(server_addr, stream_fps, sio).setup()
with edgeiq.WebcamVideoStream(cam=camera) as video_stream:
# Allow Webcam to warm up
time.sleep(2.0)
fps.start()
prevModelChoice = None
while True:
if model_choice == "q":
print('Exiting program...')
if streamer is not None:
streamer.close()
fps.stop()
return
model = models[model_choice]
if model_choice != prevModelChoice and prevModelChoice is not None:
displayRuntimeStatistics(model)
frame = video_stream.read()
frame, text = runModel(model, frame, model_choice)
streamer.send_data(frame, text)
fps.update()
if streamer.check_exit():
break
finally:
if streamer is not None:
streamer.close()
fps.stop()
print("elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("approx. FPS: {:.2f}".format(fps.compute_fps()))
print("Program Ending")
def object_detection():
"""
Oak.get_model_result can return different results based on the purpose of the model running on the camera.
This function shows how to work with object detection models.
"""
fps = edgeiq.FPS()
with edgeiq.Oak('alwaysai/mobilenet_ssd_oak') as camera, edgeiq.Streamer(
) as streamer:
fps.start()
while True:
text = ['FPS: {:2.2f}'.format(fps.compute_fps())]
frame = camera.get_frame()
result = camera.get_model_result(confidence_level=.75)
# Check for inferencing results. Oak.get_model_result is a non-blocking call and will return None when new data is not available.
if result:
frame = edgeiq.markup_image(frame, result.predictions)
text.append("Objects:")
for prediction in result.predictions:
text.append("{}: {:2.2f}%".format(
prediction.label, prediction.confidence * 100))
streamer.send_data(frame, text)
if streamer.check_exit():
break
fps.update()
print('fps = {}'.format(fps.compute_fps()))
def main():
"""Run CSI camera."""
try:
with edgeiq.JetsonVideoStream(cam=0,
rotation=edgeiq.FrameRotation.ROTATE_180,
camera_mode=edgeiq.JetsonCameraMode.
IMX219_1920x1080_30_2, display_width=640,
display_height=480) as video_stream, \
edgeiq.Streamer() as streamer:
# Allow Webcam to warm up
time.sleep(2.0)
# loop detection
while True:
frame = video_stream.read()
# Generate text to display on streamer
text = ["Jetson CSI Camera"]
streamer.send_data(frame, text)
if streamer.check_exit():
break
finally:
print("Program Ending")
def main():
obj_detect = edgeiq.ObjectDetection("alwaysai/yolo_v3")
obj_detect.load(engine=edgeiq.Engine.DNN)
video_path = f"data/inputs/{difficulty}.mp4"
stream_context = edgeiq.FileVideoStream(f"{video_path}",
play_realtime=True)
with stream_context as video_stream, edgeiq.Streamer() as streamer:
while video_stream.more():
image = video_stream.read()
results = obj_detect.detect_objects(image, confidence_level=.5)
specific_predictions = [
r for r in results.predictions if r.label == 'person'
]
res = tracker.update(specific_predictions)
image = draw_tracked_boxes(image, res)
# image = edgeiq.markup_image(image, people_predictions)
streamer.send_data(image)
def main():
obj_detect = edgeiq.ObjectDetection(
"alwaysai/ssd_mobilenet_v1_coco_2018_01_28")
obj_detect.load(engine=edgeiq.Engine.DNN)
print("Engine: {}".format(obj_detect.engine))
print("Accelerator: {}\n".format(obj_detect.accelerator))
print("Model:\n{}\n".format(obj_detect.model_id))
print("Labels:\n{}\n".format(obj_detect.labels))
image_paths = sorted(list(edgeiq.list_images("images/")))
print("Images:\n{}\n".format(image_paths))
with edgeiq.Streamer(queue_depth=len(image_paths),
inter_msg_time=3) as streamer:
for image_path in image_paths:
# Load image from disk
image = cv2.imread(image_path)
results = obj_detect.detect_objects(image, confidence_level=.5)
image = edgeiq.markup_image(image,
results.predictions,
colors=obj_detect.colors)
# Generate text to display on streamer
text = ["Model: {}".format(obj_detect.model_id)]
text.append("Inference time: {:1.3f} s".format(results.duration))
text.append("Objects:")
for prediction in results.predictions:
text.append("{}: {:2.2f}%".format(prediction.label,
prediction.confidence * 100))
streamer.send_data(image, text)
streamer.wait()
print("Program Ending")
def main(camera, use_streamer, server_addr, stream_fps):
fps = edgeiq.FPS()
try:
# initialize the streamer
if use_streamer:
streamer = edgeiq.Streamer().setup()
else:
streamer = CVClient(server_addr, stream_fps, sio, writer).setup()
with edgeiq.WebcamVideoStream(cam=camera) as video_stream:
# Allow Webcam to warm up
time.sleep(2.0)
fps.start()
# loop detection
while True:
frame = video_stream.read()
text = [""]
text.append(writer.text)
streamer.send_data(frame, text)
fps.update()
if streamer.check_exit():
break
finally:
if streamer is not None:
streamer.close()
fps.stop()
print("elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("approx. FPS: {:.2f}".format(fps.compute_fps()))
print("Program Ending")
def streamer_from(config):
# print("alwaysai_helper.py: streamer_from")
should_enable = config.get(ENABLE_STREAMER, True)
if should_enable == True:
return edgeiq.Streamer()
return None
def main():
# if you would like to test an additional model, add one to the list below:
models = ["alwaysai/mobilenet_ssd", "alwaysai/ssd_inception_v2_coco_2018_01_28"]
# if you've added a model, add a new color in as a list of tuples in BGR format
# to make visualization easier (e.g. [(B, G, R)]).
colors = [[(66, 68, 179)], [(50, 227, 62)]]
detectors = []
# load all the models (creates a new object detector for each model)
for model in models:
# start up a first object detection model
obj_detect = edgeiq.ObjectDetection(model)
obj_detect.load(engine=edgeiq.Engine.DNN)
# track the generated object detection items by storing them in detectors
detectors.append(obj_detect)
# print the details of each model to the console
print("Model:\n{}\n".format(obj_detect.model_id))
print("Engine: {}".format(obj_detect.engine))
print("Accelerator: {}\n".format(obj_detect.accelerator))
print("Labels:\n{}\n".format(obj_detect.labels))
fps = edgeiq.FPS()
try:
with edgeiq.WebcamVideoStream(cam=0) as video_stream, \
edgeiq.Streamer() as streamer:
# Allow Webcam to warm up
time.sleep(2.0)
fps.start()
# loop detection
while True:
frame = video_stream.read()
text = [""]
# gather data from the all the detectors
for i in range(0, len(detectors)):
results = detectors[i].detect_objects(
frame, confidence_level=.5)
object_frame = edgeiq.markup_image(
frame, results.predictions, show_labels=False, colors=colors[i])
# for the first frame, overwrite the input feed
if i == 0:
display_frame = object_frame
else:
# otherwise, append the new marked-up frame to the previous one
display_frame = numpy.concatenate((object_frame, display_frame))
# append each prediction
for prediction in results.predictions:
text.append(
"Model {} detects {}: {:2.2f}%".format(detectors[i].model_id,
prediction.label, prediction.confidence * 100))
# send the image frame and the predictions for both
# prediction models to the output stream
streamer.send_data(display_frame, text)
fps.update()
if streamer.check_exit():
break
finally:
fps.stop()
print("elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("approx. FPS: {:.2f}".format(fps.compute_fps()))
print("Program Ending")
def main():
obj_detect = edgeiq.ObjectDetection(
"alwaysai/ssd_mobilenet_v1_coco_2018_01_28")
obj_detect.load(engine=edgeiq.Engine.DNN)
print("Engine: {}".format(obj_detect.engine))
print("Accelerator: {}\n".format(obj_detect.accelerator))
print("Model:\n{}\n".format(obj_detect.model_id))
print("Labels:\n{}\n".format(obj_detect.labels))
image_paths = sorted(list(edgeiq.list_images("images/")))
print("Images:\n{}\n".format(image_paths))
with edgeiq.Streamer(queue_depth=len(image_paths),
inter_msg_time=3) as streamer:
for image_path in image_paths:
# Load image from disk
image = cv2.imread(image_path)
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
r, g, b = cv2.split(image)
fig = plt.figure()
axis = fig.add_subplot(1, 1, 1, projection="3d")
pixel_colors = image.reshape(
(np.shape(image)[0] * np.shape(image)[1], 3))
norm = colors.Normalize(vmin=-1., vmax=1.)
norm.autoscale(pixel_colors)
pixel_colors = norm(pixel_colors).tolist()
axis.scatter(r.flatten(),
g.flatten(),
b.flatten(),
facecolors=pixel_colors,
marker=".")
axis.set_xlabel("Red")
axis.set_ylabel("Green")
axis.set_zlabel("Blue")
plt.show()
#convert from rgb to hsv and pick out 2 shades
hsv_image = cv2.cvtColor(image, cv2.COLOR_RGB2HSV)
hsv_drot = (18, 24, 61)
hsv_lrot = (13, 203, 55)
#build the color mask
mask = cv2.inRange(hsv_image, hsv_lrot, hsv_drot)
res = cv2.bitwise_and(image, image, mask=mask)
plt.subplot(1, 2, 1)
plt.imshow(mask, cmap="gray")
plt.subplot(1, 2, 2)
plt.imshow(res)
plt.show()
#2nd layer mask, did not display
hsv_olive = (34, 32, 120)
hsv_dolive = (37, 240, 27)
mask_ol = cv2.inRange(hsv_image, hsv_olive, hsv_dolive)
res_w = cv2.bitwise_and(image, image, mask=mask_ol)
plt.subplot(1, 2, 1)
plt.imshow(mask_ol, cmap="gray")
plt.subplot(1, 2, 2)
plt.imshow(res_w)
plt.show()
#final mask
final_mask = mask + mask_ol
final_result = cv2.bitwise_and(image, image, mask=final_mask)
plt.subplot(1, 2, 1)
plt.imshow(final_mask, cmap="gray")
plt.subplot(1, 2, 2)
plt.imshow(final_result)
plt.show()
#testing .shape and typecast image
print("The type of this input is {}".format(type(image)))
print("Shape: {}".format(image.shape))
#piee
##text.append(get_colors(get_image(image_path), 4, True))
# Generate text to display on streamer
text = ["Model: {}".format(obj_detect.model_id)]
text.append("Inference time: {:1.3f} s".format(results.duration))
#need to convert from bgr to rgb
swapped_colors = swap(obj_detect.colors)
text.append("Colors printed!")
# text.append(swapped_colors)
print(swapped_colors)
# print(obj_detect.colors)
# converted = np.array([np.array(rgb) for rgb in swapped_colors]) // numpy arrays with lists (like numpy contained within itself, list of lists)
# print(converted.shape)
results = obj_detect.detect_objects(image, confidence_level=.5)
image = edgeiq.markup_image(image,
results.predictions,
colors=obj_detect.colors)
# print(rgb2hex(swapped_colors))
# print(converted)
# iterate through tuple list and convert
# for x in obj_detect.colors:
# text.append(rgb2hex(swapped_colors))
# text.append(format(x))
text.append("Objects:")
for prediction in results.predictions:
text.append("{}: {:2.2f}%".format(prediction.label,
prediction.confidence * 100))
streamer.send_data(image, text)
streamer.wait()
print("Program Ending")
def main():
obj_detect = edgeiq.ObjectDetection(
"alwaysai/mobilenet_ssd")
obj_detect.load(engine=edgeiq.Engine.DNN)
tracker = edgeiq.CentroidTracker(deregister_frames=30)
print("Loaded model:\n{}\n".format(obj_detect.model_id))
print("Engine: {}".format(obj_detect.engine))
print("Accelerator: {}\n".format(obj_detect.accelerator))
print("Labels:\n{}\n".format(obj_detect.labels))
fps = edgeiq.FPS()
try:
with edgeiq.WebcamVideoStream(cam=0) as video_stream, \
edgeiq.Streamer() as streamer:
# Allow Webcam to warm up
time.sleep(2.0)
fps.start()
prev_tracked_people = {}
logs = []
currentPeople = 0
# loop detection
while True:
frame = video_stream.read()
results = obj_detect.detect_objects(frame, confidence_level=.5)
people = edgeiq.filter_predictions_by_label(results.predictions, ['person'])
tracked_people = tracker.update(people)
people = []
for (object_id, prediction) in tracked_people.items():
new_label = 'Person {}'.format(object_id)
prediction.label = new_label
people.append(prediction)
frame = edgeiq.markup_image(
frame, people, colors=obj_detect.colors)
new_entries = set(tracked_people) - set(prev_tracked_people)
for entry in new_entries:
save_snapshot(frame, entry)
logs.append('Person {} entered'.format(entry))
currentPeople += 1
new_exits = set(prev_tracked_people) - set(tracked_people)
for exit in new_exits:
logs.append('Person {} exited'.format(exit))
currentPeople -= 1
prev_tracked_people = dict(tracked_people)
# Generate text to display on streamer
text = ["Model: {}".format(obj_detect.model_id)]
text.append(
"Inference time: {:1.3f} s".format(results.duration))
text.append("Objects:")
for prediction in people:
text.append("{}: {:2.2f}%".format(
prediction.label, prediction.confidence * 100))
text.append('Current Occupancy:')
text += str(currentPeople)
streamer.send_data(frame, text)
fps.update()
if streamer.check_exit():
break
finally:
fps.stop()
print("elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("approx. FPS: {:.2f}".format(fps.compute_fps()))
print("Program Ending")
def main():
fps = edgeiq.FPS()
# Change parameter to alwaysai/human_pose_eyecloud to run the human pose model.
with edgeiq.EyeCloud('alwaysai/mobilenet_ssd_eyecloud'
) as camera, edgeiq.Streamer() as streamer:
fps.start()
centroid_tracker = edgeiq.CentroidTracker(deregister_frames=20, max_distance=100)
number=0
while True:
text = ["FPS:{}".format(fps.compute_fps())]
frame = camera.get_frame()
#print('image sequence = {}'.format(frame.sequence_index))
result = camera.get_model_result(confidence_level=0.9)
# Check for inferencing results.
if result:
#print('model sequence = {}'.format(result.sequence_index))
text.append("Model: {}".format(camera.model_id))
if camera.model_purpose == 'PoseEstimation':
frame = result.draw_poses(frame)
text.append("Inference time: {:1.3f} s".format(result.duration))
for ind, pose in enumerate(result.poses):
text.append("Person {}".format(ind))
text.append('-' * 10)
text.append("Key Points:")
for key_point in pose.key_points:
text.append(str(key_point))
elif camera.model_purpose == 'ObjectDetection':
#移除交并比过大的检测框
if(len(result.predictions)>1):
for i,prediction in enumerate(result.predictions):
for j in range((len(result.predictions)-1),i,-1):
#print("i：%s j：%s"%(i,j))
IOU=prediction.box.compute_overlap(result.predictions[j].box)
print(IOU)
if(IOU>0.3):
#print("距离过近")
result.predictions.pop()
#划定FOV有效检测区
for i,prediction in enumerate(result.predictions):
if(prediction.box.center[0]<500 or prediction.box.center[0]>1300):
result.predictions.pop(i)
#print("x：%s y：%s"%(prediction.box.center[0],prediction.box.center[1]))
objects = centroid_tracker.update(result.predictions)
text.append("Inference time: {:1.3f} s".format(result.duration))
text.append("Objects:")
for (object_id, prediction) in objects.items():
new_label = 'person {}'.format(object_id)
if(object_id+1>number):
number=object_id+1
prediction.label = new_label
text.append("{}: {:2.2f}%".format(prediction.label, prediction.confidence * 100))
result.predictions.append(prediction)
frame = edgeiq.markup_image(frame, result.predictions)
text.append("people pass: {}".format(number))
elif camera.model_purpose == 'Classification':
if len(result.predictions) > 0:
top_prediction = result.predictions[0]
text = "Classification: {}, {:.2f}%".format(
top_prediction.label,
top_prediction.confidence * 100)
else:
text = None
cv2.putText(frame, text, (5, 25), cv2.FONT_HERSHEY_SIMPLEX,
0.4, (0, 0, 255), 2)
cv2.rectangle(frame,(500,0),(1300,1080),(255,0,0), 2)
streamer.send_data(frame, text)
if streamer.check_exit():
break
fps.update()
print('fps = {}'.format(fps.compute_fps()))
def main():
pose_estimator = edgeiq.PoseEstimation("alwaysai/human-pose")
pose_estimator.load(engine=edgeiq.Engine.DNN_OPENVINO,
accelerator=edgeiq.Accelerator.MYRIAD)
print("Loaded model:\n{}\n".format(pose_estimator.model_id))
print("Engine: {}".format(pose_estimator.engine))
print("Accelerator: {}\n".format(pose_estimator.accelerator))
fps = edgeiq.FPS()
y_letter = cv2.imread('letter_y.png')
m_letter = cv2.imread('m_letter.jpg')
c_letter = cv2.imread('c_letter.jpeg')
a_letter = cv2.imread('a_letter.jpg')
try:
with edgeiq.WebcamVideoStream(cam=0) as video_stream, \
edgeiq.Streamer() as streamer:
# Allow Webcam to warm up
time.sleep(2.0)
fps.start()
# loop detection
while True:
frame = video_stream.read()
results = pose_estimator.estimate(frame)
# Generate text to display on streamer
text = [""]
for ind, pose in enumerate(results.poses):
right_wrist_y = pose.key_points[4][1]
right_wrist_x = pose.key_points[4][0]
right_elbow_y = pose.key_points[3][1]
right_elbow_x = pose.key_points[3][0]
left_wrist_y = pose.key_points[7][1]
left_wrist_x = pose.key_points[7][0]
left_elbow_y = pose.key_points[6][1]
left_elbow_x = pose.key_points[6][0]
nose_y = pose.key_points[0][1]
nose_x = pose.key_points[0][0]
neck_y = pose.key_points[1][0]
if nose_y != -1 and neck_y != -1:
neck_distance = neck_y - nose_y
else:
neck_distance = 0
if right_wrist_y != -1 and left_wrist_y != -1 and nose_y != -1 and left_elbow_y != -1 and right_elbow_y != -1 and neck_distance > 0:
if right_wrist_y < nose_y and left_wrist_y < nose_y and right_wrist_x > right_elbow_x and left_wrist_x < left_elbow_x:
if right_wrist_y < (
nose_y -
neck_distance / 3.0) and left_wrist_y < (
nose_y - neck_distance / 3.0):
print("----------A!-------------")
overlay = edgeiq.resize(
a_letter, frame.shape[1], frame.shape[0],
False)
cv2.addWeighted(frame, 0.4, overlay, 0.6, 0,
frame)
continue
elif (nose_y - neck_distance) < right_wrist_y and (
nose_y - neck_distance) < left_wrist_y:
print("----------M!-------------")
overlay = edgeiq.resize(
m_letter, frame.shape[1], frame.shape[0],
False)
cv2.addWeighted(frame, 0.4, overlay, 0.6, 0,
frame)
continue
if right_wrist_y != -1 and left_wrist_y != -1 and nose_y != -1 and right_elbow_x and left_elbow_x and right_wrist_x and left_wrist_x:
if right_wrist_y < nose_y and left_wrist_y < nose_y and right_wrist_x < right_elbow_x and left_wrist_x > left_elbow_x:
print("----------Y!-------------")
overlay = edgeiq.resize(y_letter, frame.shape[1],
frame.shape[0], False)
cv2.addWeighted(frame, 0.4, overlay, 0.6, 0, frame)
continue
if left_wrist_x != -1 and nose_x != -1 and left_wrist_y != -1 and nose_y != -1 and right_wrist_y != -1 and nose_x != -1:
if right_wrist_x > nose_x and right_wrist_y < nose_y and left_wrist_x > nose_x:
print("----------C!-------------")
overlay = edgeiq.resize(c_letter, frame.shape[1],
frame.shape[0], False)
cv2.addWeighted(frame, 0.4, overlay, 0.6, 0, frame)
continue
streamer.send_data(results.draw_poses(frame), text)
fps.update()
if streamer.check_exit():
break
finally:
fps.stop()
print("elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("approx. FPS: {:.2f}".format(fps.compute_fps()))
print("Program Ending")
def main():
# The current frame index
frame_idx = 0
# The number of frames to skip before running detector
detect_period = 50
# if you would like to test an additional model, add one to the list below:
models = [
"alwaysai/ssd_mobilenet_v2_oidv4",
"alwaysai/ssd_inception_v2_coco_2018_01_28"
]
# include any labels that you wish to detect from any models (listed above in 'models') here in this list
detected_contraband = [
"Pen", "cell phone", "backpack", "book", "Book", "Ring binder",
"Headphones", "Calculator", "Mobile phone", "Telephone", "Microphone",
"Ipod", "Remote control"
]
# load all the models (creates a new object detector for each model)
detectors = []
for model in models:
# start up a first object detection model
obj_detect = edgeiq.ObjectDetection(model)
obj_detect.load(engine=edgeiq.Engine.DNN)
# track the generated object detection items by storing them in detectors
detectors.append(obj_detect)
# print the details of each model to the console
print("Model:\n{}\n".format(obj_detect.model_id))
print("Engine: {}".format(obj_detect.engine))
print("Accelerator: {}\n".format(obj_detect.accelerator))
print("Labels:\n{}\n".format(obj_detect.labels))
tracker = edgeiq.CorrelationTracker(max_objects=5)
fps = edgeiq.FPS()
contraband_summary = ContrabandSummary()
try:
with edgeiq.WebcamVideoStream(cam=0) as video_stream, \
edgeiq.Streamer() as streamer:
# Allow Webcam to warm up
time.sleep(2.0)
fps.start()
# loop detection
while True:
frame = video_stream.read()
predictions_to_markup = []
text = [""]
# only analyze every 'detect_period' frame (i.e. every 50th in original code)
if frame_idx % detect_period == 0:
# gather data from the all the detectors
for i in range(0, len(detectors)):
results = detectors[i].detect_objects(
frame, confidence_level=.2)
# Stop tracking old objects
if tracker.count:
tracker.stop_all()
# append each prediction
predictions = results.predictions
for prediction in predictions:
if (prediction.label.strip()
in detected_contraband):
contraband_summary.contraband_alert(
prediction.label, frame)
predictions_to_markup.append(prediction)
tracker.start(frame, prediction)
else:
# if there are objects being tracked, update the tracker with the new frame
if tracker.count:
# get the new predictions for the objects being tracked, used to markup the frame
predictions_to_markup = tracker.update(frame)
# mark up the frame with the predictions for the contraband objects
frame = edgeiq.markup_image(frame,
predictions_to_markup,
show_labels=True,
show_confidences=False,
colors=obj_detect.colors)
# send the collection of contraband detection points (string and video frame) to the streamer
text = contraband_summary.get_contraband_string()
streamer.send_data(frame, text)
frame_idx += 1
fps.update()
if streamer.check_exit():
break
finally:
fps.stop()
print("elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("approx. FPS: {:.2f}".format(fps.compute_fps()))
print("Program Ending")
def main():
# Spin up the object detector
obj_detect = edgeiq.ObjectDetection("alwaysai/" + OBJECT_DETECTION_MODEL)
obj_detect.load(engine=edgeiq.Engine.DNN_CUDA,
accelerator=edgeiq.Accelerator.NVIDIA)
print("Engine: {}".format(obj_detect.engine))
print("Accelerator: {}\n".format(obj_detect.accelerator))
print("Model:\n{}\n".format(obj_detect.model_id))
print("Labels:\n{}\n".format(obj_detect.labels))
# Prepare to track frames per second calculations
fps = edgeiq.FPS()
# Load any prior instance of the tracker, otherwise spin up a new one
centroid_tracker = file_manager.load(
CENTROID_TRACKER,
edgeiq.CentroidTracker(deregister_frames=TRACKER_DEREGISTER_FRAMES,
max_distance=TRACKER_MAX_DISTANCE))
# Load any prior instance of the metrics data, otherwise start a new one
metrics = file_manager.load(METRICS_MANAGER,
metrics_manager.MetricsManager())
try:
if IP_CAMERA_FEED is not None:
stream_details = edgeiq.IPVideoStream(IP_CAMERA_FEED)
else:
stream_details = edgeiq.WebcamVideoStream(cam=0)
with stream_details as video_stream, \
edgeiq.Streamer() as streamer:
# Allow Webcam to warm up
time.sleep(2.0)
fps.start()
# Loop detection and centroid tracker
while True:
metrics.newLoop()
frame = video_stream.read()
results = obj_detect.detect_objects(
frame, confidence_level=DETECT_CONFIDENCE_THRESHOLD)
# Ignore detections of anything other than people
filter = edgeiq.filter_predictions_by_label(
results.predictions, ['person'])
# Adding info for streamer display
text = ["Model: {}".format(obj_detect.model_id)]
text.append("Inference time: {:1.3f} s".format(
results.duration))
text.append("People currently detected:")
objects = centroid_tracker.update(filter)
# Store active predictions for just this loop
predictions = []
# Store the active object ids for just this loop
if len(objects.items()) == 0:
# No people detected
text.append("-- NONE")
for (object_id, prediction) in objects.items():
metrics.addTimeFor(object_id)
timeForId = metrics.timeForId(object_id)
# Correcting for fact that index 0 is first object in an array
idAdjusted = object_id + 1
# Display text with bounding box in video
new_label = "Person {i} | {t} sec".format(i=idAdjusted,
t=timeForId)
prediction.label = new_label
text.append(new_label)
predictions.append(prediction)
# Add metrics to text going to streamer
m = metrics.currentMetrics()
text.append("") # Spacing
text.append("Total people seen: {}".format(m["count"]))
text.append("Total time: {} sec".format(m["total"]))
text.append("Average time: {0:.1f} sec".format(m["avg"]))
text.append("Longest individual time: {} sec".format(m["max"]))
# Update output streamer
frame = edgeiq.markup_image(frame, predictions)
streamer.send_data(frame, text)
fps.update()
if streamer.check_exit():
break
finally:
fps.stop()
# TODO: Update to save every few seconds in case a crash occurs
file_manager.save(metrics, METRICS_MANAGER)
file_manager.save(centroid_tracker, CENTROID_TRACKER)
print("elapsed time: {:.2f}".format(fps.get_elapsed_seconds()))
print("approx. FPS: {:.2f}".format(fps.compute_fps()))
print("Program Ending")
