O tekinolosi e iloa ai mea faitino a komepiuta e taua tele mo le tele o talosaga. Matou te faʻaaogaina i robotics, masini mataʻituina, taʻavale a le tagata lava ia, ma le tele o isi vaega. O le mea lea, e mafai ona tatou maua ma iloa nisi mea i se ata poʻo se vitio.
O se tasi e sili ona lauiloa mea e iloagofie ai algorithms o le YOLO (Na'o E Va'ai Fa'atasi) seti o fa'ata'ita'iga. O nei faʻataʻitaʻiga e faia e Ultralytics LLC.
Ole lomiga lata mai ole fa'asologa ole YOLOv5. Ma, o le faʻataʻitaʻiga faʻamatalaga sili ona vave ma sili ona saʻo i luga o le maketi. O le gafatia o le faʻataʻitaʻiga e faʻasalalau i faʻamatalaga fou ua matua faʻaleleia. E le gata i lea, o loʻo i ai le tele o mea e faʻaleleia atili ai nai lo le faʻasologa muamua.
O le YOLOv5 e lelei tele mo fa'aoga taimi moni talu ai e mafai ona fa'agasolo ata i se fua faatatau e o'o atu i le 1000 fa'avaa i le sekone i luga ole GPU tasi.
I lenei tusiga, o le a matou faʻalauiloaina le YOLOv5 ma faʻasalalau faʻamatalaga o ona vaega o faʻaoga.
Malaga a YOLO: Mai le YOLO i le YOLOv5
Iosefa Redmon et al. na muamua faʻafeiloaʻi le YOLO, o se seti o faʻataʻitaʻiga faʻataʻitaʻiga, i le 2016. O le faʻataʻitaʻiga muamua o le YOLO e mafai ona iloa ai mea i le taimi moni. Ae ui i lea, e maualalo le saʻo pe a faʻatusatusa i isi faʻataʻitaʻiga i lena taimi.
E tele fa'afouga fa'afouga o le YOLO na fa'asa'oloto i le gasologa o tausaga. Ma mulimuli ane, na faia e Ultralytics LLC le lomiga fou o le YOLO series, YOLOv5.
O le YOLOv5 o le fa'ata'ita'iga sili ona sa'o ma sili ona vave fa'ailoa mea o lo'o avanoa nei.
Taua Taua
Pusa Taula
YOLOv5 va'ai pusa fusifusia mo mea faitino i se ata e fa'aaoga ai pusa taula. O le fa'ata'ita'iga o lo'o va'ai po'o fea o le tele o atigipusa na mua'i fa'auigaina ma fua fa'atatau eseese e sili ona fetaui ma mea o lo'o i le ata e fa'aaoga ai pusa taula. O pusa ia na muai faʻamalamalamaina.
Ma, latou te mafaia e YOLOv5 ona iloa ma suʻe mea i totonu o se ata ma le saʻo.
Fa'aopoopo fa'amaumauga a Mosaic
A'o a'oa'oga, e fa'aaogaina e le YOLOv5 se metotia e ta'ua o le mosaic fa'aopoopoga fa'amatalaga. Ina ia atia'e ata fou a'oa'oga, o la matou fa'ata'ita'iga e tu'ufa'atasia fa'atasi ai ni pa'i o ni ata. O se taunuuga, o le faʻataʻitaʻiga e sili atu ona faʻaleleia ma faʻalagolago. O le mea lea, e mafai ona faʻasalalau i faʻamatalaga fou ma faʻaitiitia le faʻaogaina.
Ose Pipeline A'oa'oga Tulaga
O se paipa a'oa'oga tulaga ese e fa'afefiloi le vaavaaia ma aʻoaʻoga le faʻatautaia e faʻaaoga.
O lea la, o le fa'ata'ita'iga e a'oa'oina mai se fa'ata'ita'iga la'ititi ma fa'aoga lelei mea e le fa'ailogaina. Ole mea lea e fa'ateleina ai le fa'atinoga o le fa'ata'ita'iga ma fa'ateleina lona gafatia e fa'alauteleina i mea fou.
Laega e totoe ma e le o totoe
O le fausaga a le YOLOv5 o lo'o tu'ufa'atasia ai laulau o lo'o totoe ma e le o totoe. E ala i le fa'ataga o fa'alili e fa'asolo atu i luga o fa'apalapala, o fa'a'aga o totoe e fesoasoani i le fa'ata'ita'iga i le a'oa'oina o vaega faigata. E le gata i lea, o laupepa e le o totoe e maua ai le faʻataʻitaʻiga ma se malamalama sili atu o le ata faʻaoga. O le i'uga, YOLOv5 e mafai ona fa'agaoioi ma le sa'o ma lelei.
Fa'afefea ona fa'aoga YOLOv5
oʻo
YOLOv5 faʻapipiʻi e mafai ona faʻamaeʻa vave e faʻaaoga ai le pip. Pip o le Python package manager. O faiga masani mo le faʻapipiʻiina o le YOLOv5 e faʻapea:
1- Faʻapipiʻi PyTorch: Ona o le YOLOv5 e faʻavae i luga o le PyTorch framework, e tatau ona e faʻapipiʻi muamua PyTorch.
pip install torch torchvision
2. Faʻapipiʻi le CUDA: E tatau ona e faʻapipiʻi le CUDA pe afai e te manaʻo e taʻavale YOLOv5 i luga ole GPU.
3. Faʻapipiʻi YOLOv5: A maeʻa ona faʻatulagaina PyTorch ma CUDA, faʻaaoga le poloaiga lenei e sii mai ai YOLOv5.
pip install yolov5
4-Ile mulimuli i le faʻapipiʻiina o le YOLOv5, e tatau ona e siiina le mamafa muamua. O loʻo maua le mamafa muamua i le Ultralytics GitHub repo.
Alu i le "mamafa" vaega o le upega tafaʻilagi e ala i le taʻavale i lalo. E mafai ona e sii maia le mamafa ua uma ona a'oa'oina mai le lisi e mafai ona e mauaina iinei.
5. Filifili le mamafa ua uma ona aʻoaʻoina ma sili ona fetaui ma lau faʻaoga faʻaoga. Ole fa'amaumauga po'o le YOLOv5 fa'apitoa na a'oa'oina le mamafa e mafai ona fa'aoga e fa'aiti ai le lisi.
6- A maeʻa ona filifili le mamafa saʻo, filifili le mamafa e ala i le kilikiina o le "Download" button i tafatafa. O le mamafa o le a avanoa mo le download e pei o. pt faila.
7- Faʻaliliuina le mamafa o loʻo siiina i le lisi. O i'inei o le a fa'agaioi ai lau fa'amatalaga su'esu'e.
8- I le taimi nei, e mafai ona e faʻatautaia le suʻesuʻeina o mea i luga o au ata poʻo vitio e faʻaaoga ai le mamafa na muaʻi aʻoaʻoina i lau tusitusiga.
Saunia Faamatalaga
E tatau ona e faia gaioiga nei e saunia ai faʻamatalaga mo le faʻaaogaina ma le YOLOv5:
1. Fa'apotopoto fa'amaumauga: O le la'asaga muamua o le fa'aputuina lea o fa'amaumauga o ata po'o vitio e te mana'omia mea e sailia. O mea e te mana'o e iloa e tatau ona iai i ata po'o vitio.
2- Fa'asologa o fa'amaumauga: E mafai ona e fa'aulufale mai ata i totonu o lau tusitusiga pe a e fa'aaogaina. E tatau ona e liliu se vitio i se faasologa o ata pe afai e te fuafua e faʻaaoga se tasi. E mafai ona e aveese fa'avaa mai se ata e fa'aoga ai se faletusi pei o OpenCV.
import cv2
img = cv2.imread('path/to/image')
Faatasi ai ma le OpenCV faletusi, e mafai ona e faʻaogaina le poloaiga lenei e liliu ai se vitio i se faasologa o ata:
import cv2
cap = cv2.VideoCapture('path/to/video')
while True:
ret, frame = cap.read()
if not ret:
break
cv2.imshow('frame', frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cap.release()
cv2.destroyAllWindows()
3. Fa'ailoga fa'amaumauga: E tatau ona e fa'ailogaina fa'amaumauga pe afai o lo'o e fa'aogaina lau fa'amaumauga. Tusia pusa fusia faataamilo i aitema e te manao e faailoa i faavaa taitasi o se ata. O le fa'agasologa o le fa'ailogaina o fa'amaumauga. E mafai ona e fa'aogaina ni mea faigaluega e fesoasoani ia te oe i lenei fa'agaioiga, e aofia ai le LabelImg ma RectLabel.
4- E tatau ona e vaevae faʻamaumauga i aʻoaʻoga ma seti suʻega pe a uma ona e faʻailogaina. E taua tele lenei mea mo le iloiloina o le lelei o lau faʻataʻitaʻiga.
5. Mulimuli ane, atonu e te mana'omia le mua'i fa'agaioiina o fa'amaumauga a'o le'i a'oa'oina po'o le su'ega. O lenei mea e mafai ona aofia ai le faʻavasegaina o ata poʻo vitio, faʻavasegaina o tau pika, poʻo le faʻaogaina o metotia mo le faʻaopoopoga o faʻamatalaga.
A maeʻa nei laasaga, ua saunia au faʻamatalaga.
Fa'agasolo le fa'amatalaga su'esu'e
O se fa'ata'ita'iga lea o se fa'amatalaga e su'e ai se ata ma su'e mea.
import yolov5
import cv2
# Pre-trained weights should be loaded.
weights = 'path/to/weights.pt'
# Set the detection confidence level
conf_thres = 0.5
# Set the Non-Maxima Suppression (NMS) threshold
nms_thres = 0.5
# Create the detector object
detector = yolov5.YOLOv5(weights, conf_thres, nms_thres)
# Load the image
img = cv2.imread('path/to/image')
# Perform object detection
detections = detector.detect(img)
# Print the detections
for x1, y1, x2, y2, conf, cls_conf, cls_pred in detections:
print("Object:", classes[int(cls_pred)])
print("Confidence:", conf)
print("Bounding box:", (x1, y1, x2, y2))
Fale-gaosi
O le taofiofia e le sili ona maualuga o se tasi lea o auala e masani ona faʻaaogaina i le faʻaogaina o mea (NMS). Matou te faʻaogaina le NMS e faʻaumatia pusa faʻapipiʻi faʻapipiʻi mo le mea lava e tasi. Ina ia faʻatino NMS i luga o suʻesuʻega, e mafai ona matou faʻaogaina le OpenCV library's cv2.dnn.NMSBoxes() method.
O se fa'ata'ita'iga lea o le fa'aogaina o su'esu'ega pe a mae'a fa'aogaina le NMS.
import cv2
# Perform Non-Maxima Suppression (NMS)
fa'ailoga = cv2.dnn.NMSBoxes(su'esu'ega, fa'amaoni, conf_thres, nms_thres)
Faʻaaliga vaaia
I le tulaga o le vaʻaia, e mafai ona tatou toe faʻaogaina se faletusi e pei o OpenCV. E mafai ona tatou fa'aali pusa fusi fa'ata'amilo i mea na maua i luga o le ata po'o le vitio. Ina ia tusia pusa fusifusia o le ata, faaaoga le cv2.rectangle() method. O le auala lenei e va'ai ai fa'amatalaga i le ata muamua:
faaulufale mai cv2
# Draw the bounding boxes on the image
aua o aʻu i faʻamatalaga:
i = i[0]
x1, y1, x2, y2 = detections[i][0], detections[i][1], detections[i][2], detections[i][3]
cv2.rectangle(img, (x1, y1), (x2, y2), (255, 0, 0), 2)
cv2.putText(img, classes[class_ids[i]], (x1, y1), cv2.FONT_HERSHEY_SIMPLEX, 1, (255, 255, 255), 2)
# Show the image
cv2.imshow("Object Detection", img)
cv2
talosaga
YOLOv5 o se faʻataʻitaʻiga faʻataʻitaʻiga malosi. O le mea lea, e mafai ona tatou faʻaaogaina i le tele o faʻaaliga moni o le lalolagi. O se tasi o faʻaoga sili ona lauiloa o loʻo i totonu o taavale taʻavale. E mafai e le YOLOv5 ona iloa mea i le taimi moni e pei o taavale ma moli auala.
I faiga mata'ituina, e mafai ona tatou fa'aogaina le YOLOv5 e iloa ma siaki ai mea i ata vitio ola. E le gata i lea, o le YOLOv5 e mafai ona avea ma aseta sili i robotics. E mafai ona fesoasoani i robots e iloa ma malamalama i o latou siosiomaga. E taua tele lenei mea mo gaioiga e pei o le faʻaogaina ma le faʻaogaina.
E mafai fo'i ona fa'aoga YOLOv5 i so'o se alamanuia e mana'omia le su'eina o mea, e pei o fa'atau, ta'aloga, foma'i, ma le saogalemu.
iʻuga
Ma le mea mulimuli, YOLOv5 o le faʻamatalaga sili ona lata mai ma faʻapitoa o le YOLO aiga o mea e sailia Fa'ata'ita'iga
. E le gata i lea, e sa'o le fa'apea o le fa'ata'ita'iga sili ona sa'o o le su'esu'eina o mea e maua. Fa'afetai i lona maualuga sa'o ma le saoasaoa, e mafai ona e filifilia ma le saogalemu mo au galuega fa'atino.
Resky Agus
Ou te faia se tusi talaaga muamua e uiga i le taʻavale suʻesuʻe ma yolov5 ma o lenei upega tafaʻilagi fesoasoani ia te aʻu e suʻe faʻamatalaga e uiga i lena mea.
Ou te matua fiafia lava i AI.
afai e te mafaia e tele saʻu fesili e uiga i AI atonu e mafai ona e fesoasoani mai ia te aʻu
Faafetai