我的问题是：
代码：
#hoangAI
#DIKY STACKOVERFLOW JSEM SE NAUCIL OOP V PYTHONU DIKY <3
#TENTO SCRIPT JE PRO DETEKCI OBJEKTU SPECIFIKOVANYCH V CLASSES[]
#omlouvam se za anglicke nazvy promennych, prislo mi blby to jmenovat cesky
# importujeme vsechny potrebne knihovny(dekuji moc stackoverflow :3)
from imutils.video import VideoStream
from imutils.video import FPS
import numpy as np
import argparse
import imutils
import time
import cv2
import random
#prikazove argumenty
ap = argparse.ArgumentParser()
ap.add_argument("-p", "--prototxt", required=True,
help="pridej k argumentu -p path k prototxt souboru")
ap.add_argument("-m", "--model", required=True,
help="pridej path k trenovacimu modelu")
ap.add_argument("-c", "--confidence", type=float, default=0.2,
help="minimalni sance (na odstraneni tzv slabych detekci")
args = vars(ap.parse_args())
#"nacteni" hoangAI :D
print("Loading hoangAI(TM) V.1.0.0")
time.sleep(2.0)
# nacteme si list stitku veci, ktere byly pouzity pri detekcnim treningu hoangAI, a pak nahodne vybereme barvu pro objektove ramecky
CLASSES = ["background", "aeroplane", "bicycle", "bird", "boat",
"bottle", "bus", "car", "cat", "chair", "cow", "diningtable",
"dog", "horse", "motorbike", "person", "pottedplant", "sheep",
"sofa", "train", "tvmonitor"]
CLASSES_RANDOM = CLASSES[random.randint(0,len(CLASSES) - 1)]
print("vygoogli si" + CLASSES_RANDOM)
COLORS = np.random.uniform(0, 255, size=(len(CLASSES), 3))
# nacteme si serializovany model caffe
print("[INFO] nacitani modelu...")
net = cv2.dnn.readNetFromCaffe(args["prototxt"], args["model"])
# zacatek video snimani
# a take zacneme pocitat fps
print("[INFO] video snimani zacina...TED")
vs = VideoStream(src=0).start()
time.sleep(2.0)
fps = FPS().start()
# cyklus/smycka pres ramecek vystupu videa
while True:
# vezmeme ramecek z threadovaneho video vystupu a omezime sirku ramecku na 400 pixelu(aby to nebylo moc narocne na vykon)
frame = vs.read()
frame = imutils.resize(frame, width=400)
# vezmem rozmery ramecku a prevedeme do blobu(objektu)
(h, w) = frame.shape[:2]
blob = cv2.dnn.blobFromImage(cv2.resize(frame, (300, 300)),
0.007843, (300, 300), 127.5)
# objekt(blob) projde siti aby ziskal detekce a predpovedi
net.setInput(blob)
detections = net.forward()
# cyklus/smycka pres detekce
for i in np.arange(0, detections.shape[2]):
# vytahnout confidence (tedy jak si je svym uhodnutim pocitac jisty) spojeny s predpovedi
confidence = detections[0, 0, i, 2]
# odstranime "slabe" detekce s tim ze `confidence` (tedy jak si je svym uhodnutim pocitac jisty) je vetsi nez minimalni confidence uvedeny v argumentaci pri spusteni
if confidence > args["confidence"]:
#vytahneme index ze stitku tridy
# `detections`, a pak vypocitame souradnice (x, y)
# pro ohranicujici ramecek objektu
idx = int(detections[0, 0, i, 1])
box = detections[0, 0, i, 3:7] * np.array([w, h, w, h])
(startX, startY, endX, endY) = box.astype("int")
# vykresleni "uhodnuti" pocitace
label = "{}: {:.2f}%".format(CLASSES[idx],
confidence * 100)
cv2.rectangle(frame, (startX, startY), (endX, endY),
COLORS[idx], 2)
y = startY - 15 if startY - 15 > 15 else startY + 15
cv2.putText(frame, label, (startX, y),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, COLORS[idx], 2)
# vystup z kamery a z detekcniho programu
cv2.imshow("detekce veci :DDDDDDDDDDDDDDD", frame)
key = cv2.waitKey(1) & 0xFF
# pokud je stiskle tlacitko "q" (funkce ord, aby jsme nepotrebovali input())
if key == ord("q"):
break
#funkce ktera aktualizuje fps
fps.update()
# ukonceni pocitani fps a ukonceni casovace (pouze pro test jestli to funguje)
fps.stop()
print("[INFO] cas od startu: {:.2f}".format(fps.elapsed()))
print("[INFO] prum. FPS: {:.2f}".format(fps.fps()))
# konec programu
cv2.destroyAllWindows()
vs.stop()
答案 0 :(得分：-2)
这是另一个线程中带有蜂鸣器的示例：
from threading import Event, Thread
from imutils.video import FPS
import numpy as np
import argparse
import imutils
import time
import cv2
import random
program_ending_event = Event()
buzz = False
def buzzer_script():
program_ending_event.wait()
if buzz:
pass
#Here you put your logic for the buzzer or whathave you
#Rember to return from this or the program will not exit
buzz_thread = Thread(target=buzzer_script)
buzz_thread.start()
ap = argparse.ArgumentParser()
ap.add_argument("-p", "--prototxt", required=True,
help="pridej k argumentu -p path k prototxt souboru")
ap.add_argument("-m", "--model", required=True,
help="pridej path k trenovacimu modelu")
ap.add_argument("-c", "--confidence", type=float, default=0.2,
help="minimalni sance (na odstraneni tzv slabych detekci")
args = vars(ap.parse_args())
# "nacteni" hoangAI :D
print("Loading hoangAI(TM) V.1.0.0")
time.sleep(2.0)
# nacteme si list stitku veci, ktere byly pouzity pri detekcnim treningu hoangAI, a pak nahodne vybereme barvu pro objektove ramecky
CLASSES = ["background", "aeroplane", "bicycle", "bird", "boat",
"bottle", "bus", "car", "cat", "chair", "cow", "diningtable",
"dog", "horse", "motorbike", "person", "pottedplant", "sheep",
"sofa", "train", "tvmonitor"]
CLASSES_RANDOM = CLASSES[random.randint(0, len(CLASSES) - 1)]
print("vygoogli si" + CLASSES_RANDOM)
COLORS = np.random.uniform(0, 255, size=(len(CLASSES), 3))
# nacteme si serializovany model caffe
print("[INFO] nacitani modelu...")
net = cv2.dnn.readNetFromCaffe(args["prototxt"], args["model"])
# zacatek video snimani
# a take zacneme pocitat fps
print("[INFO] video snimani zacina...TED")
vs = VideoStream(src=0).start()
time.sleep(2.0)
fps = FPS().start()
# cyklus/smycka pres ramecek vystupu videa
while True:
# vezmeme ramecek z threadovaneho video vystupu a omezime sirku ramecku na 400 pixelu(aby to nebylo moc narocne na vykon)
frame = vs.read()
frame = imutils.resize(frame, width=400)
# vezmem rozmery ramecku a prevedeme do blobu(objektu)
(h, w) = frame.shape[:2]
blob = cv2.dnn.blobFromImage(cv2.resize(frame, (300, 300)),
0.007843, (300, 300), 127.5)
# objekt(blob) projde siti aby ziskal detekce a predpovedi
net.setInput(blob)
detections = net.forward()
# cyklus/smycka pres detekce
detected_class = (0, 0)
for i in np.arange(0, detections.shape[2]):
# vytahnout confidence (tedy jak si je svym uhodnutim pocitac jisty) spojeny s predpovedi
confidence = detections[0, 0, i, 2]
# odstranime "slabe" detekce s tim ze `confidence` (tedy jak si je svym uhodnutim pocitac jisty) je vetsi nez minimalni confidence uvedeny v argumentaci pri spusteni
if confidence > args["confidence"]:
# vytahneme index ze stitku tridy
# `detections`, a pak vypocitame souradnice (x, y)
# pro ohranicujici ramecek objektu
idx = int(detections[0, 0, i, 1])
# Check if new class has the highest probability <--------------- ADDED
if confidence > detected_class[0]:
detected_class = (confidence, CLASSES[idx])
box = detections[0, 0, i, 3:7] * np.array([w, h, w, h])
(startX, startY, endX, endY) = box.astype("int")
# vykresleni "uhodnuti" pocitace
label = "{}: {:.2f}%".format(CLASSES[idx],
confidence * 100)
cv2.rectangle(frame, (startX, startY), (endX, endY),
COLORS[idx], 2)
y = startY - 15 if startY - 15 > 15 else startY + 15
cv2.putText(frame, label, (startX, y),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, COLORS[idx], 2)
if detected_class[1] == CLASSES_RANDOM:
buzz = True
break
# vystup z kamery a z detekcniho programu
cv2.imshow("detekce veci :DDDDDDDDDDDDDDD", frame)
key = cv2.waitKey(1) & 0xFF
# pokud je stiskle tlacitko "q" (funkce ord, aby jsme nepotrebovali input())
if key == ord("q"):
break
# funkce ktera aktualizuje fps
fps.update()
program_ending_event.set() #<----------- SIGNAL PROGRAM END
# ukonceni pocitani fps a ukonceni casovace (pouze pro test jestli to funguje)
fps.stop()