def show_inference(model, image_path):
image_np = np.array(image_path)
lane_image = copy.deepcopy(image_np)
height, width, channel = image_np.shape
output_dict = functions.get_dict(dashPointer, model, image_np)
confidencesCars, boxesCars, confidencesLights, boxesLights, confidencesPersons, boxesPersons = functions.findBoxes(
image_np, output_dict)
indexesLights = cv2.dnn.NMSBoxes(boxesLights, confidencesLights, 0.5, 0.4)
indexesCars = cv2.dnn.NMSBoxes(boxesCars, confidencesCars, 0.5, 0.4)
indexesPersons = cv2.dnn.NMSBoxes(boxesPersons, confidencesPersons, 0.5,
0.4)
image_np = signalDetection_utils.signalDetection(indexesLights,
boxesLights, image_np)
image_np = tracking_utils.tracking(indexesCars, boxesCars, image_np)
image_np = estimate_collide_utils.estimate_collide(indexesCars, boxesCars,
image_np)
image_np = estimate_stepping_utils.estimate_stepping(
indexesPersons, boxesPersons, image_np)
cv2.putText(image_np, "DAY", (width - 200, 50), font, 2, (167, 133, 0), 2,
cv2.LINE_AA)
image_np = lane_detection_utils.draw_lines(lanePointer, lane_image,
image_np)
# lane_detection_utils.all_lines(lanePointer , lane_image , image_np)
cv2.imshow("finally", image_np)
def show_inference(dashPointer, lanePointer, frame):
global flagPerson, areaPerson, areaDetails
global crash_count_frames
global signalCounter, flagSignal
global prev_frame, number
global flagLanes
image_np = np.array(frame)
lane_image = copy.deepcopy(image_np)
height, width, channel = image_np.shape
output_dict = functions.get_dict(dashPointer, detection_model, image_np)
confidencesCars, boxesCars, confidencesLights, boxesLights, confidencesPersons, boxesPersons = functions.findBoxes(
image_np, output_dict)
indexesLights = cv2.dnn.NMSBoxes(boxesLights, confidencesLights, 0.5, 0.4)
indexesCars = cv2.dnn.NMSBoxes(boxesCars, confidencesCars, 0.5, 0.4)
indexesPersons = cv2.dnn.NMSBoxes(boxesPersons, confidencesPersons, 0.5,
0.4)
image_np, signalCounter, flagSignal = signalDetection_utils.signalDetection(
indexesLights, boxesLights, image_np, signalCounter, flagSignal)
image_np, prev_frame, number = tracking_utils.tracking(
indexesCars, boxesCars, image_np, prev_frame, number)
image_np, crash_count_frames = estimate_collide_utils.estimate_collide(
indexesCars, boxesCars, image_np, crash_count_frames)
image_np, flagPerson, areaPerson, areaDetails = estimate_stepping_utils.estimate_stepping(
indexesPersons, boxesPersons, image_np, flagPerson, areaPerson,
areaDetails)
cv2.putText(image_np, "DAY", (50, 90), font, 2, (167, 133, 0), 2,
cv2.LINE_AA)
image_np, flagLanes = lane_detection_utils.draw_lines(
lanePointer, dashPointer, lane_image, image_np, flagLanes)
# lane_detection_utils.all_lines(lanePointer , lane_image , image_np)
cv2.imshow("Final result", image_np)
def yolo_infer(dashPointer, lanePointer, frame):
global flagPerson, areaPerson, areaDetails
global crash_count_frames
global signalCounter, flagSignal
global prev_frame, number
global flagLanes
image_np = np.array(frame)
lane_image = copy.deepcopy(image_np)
height, width, channel = image_np.shape
mask = 255 * np.ones_like(image_np)
vertices = np.array(dashPointer, np.int32)
cv2.fillPoly(mask, [vertices], [0, 0, 0])
image_np = cv2.bitwise_and(image_np, mask)
# Detecting objects
blob = cv2.dnn.blobFromImage(image_np,
0.00392, (416, 416), (0, 0, 0),
True,
crop=False)
net.setInput(blob)
outs = net.forward(output_layers)
confidencesCars, boxesCars = [], []
confidencesLights, boxesLights = [], []
confidencesPersons, boxesPersons = [], []
for out in outs:
for detection in out:
scores = detection[5:]
class_id = np.argmax(scores)
if (class_id == 2 or class_id == 5 or class_id == 7):
scr = scores[class_id]
center_x = detection[0] * width
center_y = detection[1] * height
w = detection[2] * width
h = detection[3] * height
xmin = center_x - w / 2
ymin = center_y - h / 2
if (w * h >= 800):
confidencesCars.append(float(scr))
boxesCars.append([int(xmin), int(ymin), int(w), int(h)])
elif class_id == 0:
# print(scores)
scr = scores[class_id]
center_x = detection[0] * width
center_y = detection[1] * height
w = detection[2] * width
h = detection[3] * height
xmin = center_x - w / 2
ymin = center_y - h / 2
confidencesPersons.append(float(scr))
boxesPersons.append([int(xmin), int(ymin), int(w), int(h)])
elif class_id == 9:
scr = scores[class_id]
center_x = detection[0] * width
center_y = detection[1] * height
w = detection[2] * width
h = detection[3] * height
xmin = center_x - w / 2
ymin = center_y - h / 2
confidencesLights.append(float(scr))
boxesLights.append([int(xmin), int(ymin), int(w), int(h)])
indexesLights = cv2.dnn.NMSBoxes(boxesLights, confidencesLights, 0.5, 0.4)
indexesCars = cv2.dnn.NMSBoxes(boxesCars, confidencesCars, 0.5, 0.4)
indexesPersons = cv2.dnn.NMSBoxes(boxesPersons, confidencesPersons, 0.5,
0.4)
image_np, signalCounter, flagSignal = signalDetection_utils.signalDetection(
indexesLights, boxesLights, image_np, signalCounter, flagSignal)
image_np, prev_frame, number = tracking_utils.tracking(
indexesCars, boxesCars, image_np, prev_frame, number)
image_np, crash_count_frames = estimate_collide_utils.estimate_collide(
indexesCars, boxesCars, image_np, crash_count_frames)
image_np, flagPerson, areaPerson, areaDetails = estimate_stepping_utils.estimate_stepping(
indexesPersons, boxesPersons, image_np, flagPerson, areaPerson,
areaDetails)
cv2.putText(image_np, "DAY", (width - 200, 50), font, 2, (167, 133, 0), 2,
cv2.LINE_AA)
image_np, flagLanes = lane_detection_utils.draw_lines(
lanePointer, dashPointer, lane_image, image_np, flagLanes)
cv2.imshow("finally", image_np)