Ubuchwepheshe bokubona into yekhompyutha bubalulekile ezinhlelweni eziningi zokusebenza. Siyisebenzisa kumarobhothi, ezintweni zokugada, ezimotweni ezizishayelayo, nakwezinye izindawo eziningi. Ngakho-ke, sithola futhi siqaphele izinto ezithile esithombeni noma kuvidiyo.
Omunye owaziwa kakhulu I-algorithms yokuhlonza into yi-YOLO (Ubheka Kanye Kuphela) isethi yamamodeli. Lawa mamodeli adalwe ngu I-Ultralytics LLC.
Inguqulo yakamuva yalolu chungechunge yi-YOLOv5. Futhi, iyimodeli esheshayo nenembe kakhulu yokuhlonza into emakethe. Amandla emodeli okukhiqiza idatha entsha athuthukiswe kakhulu. Futhi, iqukethe izici eziningi eziyenza isebenze kangcono kunokuphindaphinda kwangaphambilini.
I-YOLOv5 inhle ezinhlelweni zesikhathi sangempela njengoba ikwazi ukucubungula izithombe ngenani elingafika kozimele abangu-1000 ngesekhondi ku-GPU eyodwa.
Kulesi sihloko, sizokwethula i-YOLOv5 futhi sidlule imininingwane yezindawo zayo zokusebenza.
Uhambo lwe-YOLO: Ukusuka e-YOLO ukuya ku-YOLOv5
UJoseph Redmon et al. ekuqaleni yethula i-YOLO, isethi yamamodeli okuhlonza into, ngo-2016. Imodeli yokuqala ye-YOLO ibingakwazi ukuhlonza izinto ngesikhathi sangempela. Nokho, yayinokunemba okuphansi uma iqhathaniswa namanye amamodeli ngaleso sikhathi.
Izinguqulo ezimbalwa ezithuthukisiwe ze-YOLO zakhululwa phakathi neminyaka. Ekugcineni, i-Ultralytics LLC idale uhlelo olusha kakhulu lochungechunge lwe-YOLO, YOLOv5.
I-YOLOv5 iyimodeli yokuhlonza into enembe kakhulu neshesha kunazo zonke etholakalayo njengamanje.
Izici ezibalulekile
Amabhokisi Anchor
I-YOLOv5 ibikezela amabhokisi ahlanganisayo ezinto ezisesithombeni kusetshenziswa amabhokisi okusetshenzwa kulengwa. Imodeli ibikezela ukuthi yimaphi amabhokisi amaningi achazwe ngaphambilini ane-aspect ratio ehlukahlukene afanelana kangcono nento esesithombeni kusetshenziswa amabhokisi okubamba ihange. Lawa amabhokisi achazwe ngaphambilini.
Futhi, zenza i-YOLOv5 ibone futhi ithole izinto esithombeni ngokunemba.
Ukwengezwa kwedatha kaMose
Lapho uqeqeshwa, i-YOLOv5 isebenzisa indlela eyaziwa ngokuthi yi-mosaic ukwengezwa kwedatha. Ukwakha izithombe zokuqeqesha ezintsha, imodeli yethu ihlanganisa ngokungahleliwe amapeshi ezithombe ezimbalwa. Ngenxa yalokho, imodeli iba namandla futhi ithembekile. Ngakho-ke, ifinyelela kudatha entsha futhi yehlise ukufaka ngokweqile.
Ipayipi Lokuqeqesha Eliyingqayizivele
Ipayipi lokuqeqesha eliyingqayizivele elixuba eligadiwe kanye ukufunda okungajwayelekile isetshenziswa.
Ngakho, imodeli ifunda kusampula encane futhi isebenzise okokufaka okungalebulanga ngempumelelo. Lokhu kuthuthukisa ukusebenza kwemodeli futhi kuthuthukisa amandla ayo okuhlanganisa okokufaka okusha.
Izendlalelo eziyinsalela nezingezona ezosalayo
I-YOLOv5's architecture ihlanganisa izendlalelo eziyinsalela nezingezona izinsalela. Ngokuvumela ama-gradient ageleze kuzo zonke izendlalelo, izendlalelo ezisele zisiza imodeli ekufundeni izici ezinzima. Futhi, izendlalelo ezingezona izinsalela zinikeza imodeli ukuqonda okubanzi kwesithombe sokufaka. Njengomphumela, i-YOLOv5 ingasebenza ngokunembe nangempumelelo.
Isetshenziswa kanjani i-YOLOv5
Ukufakwa
Ukufakwa kwe-YOLOv5 kungase kuqedwe ngokushesha kusetshenziswa i-pip. UPip ungumphathi wephakheji wePython. Izinqubo ezijwayelekile zokufaka i-YOLOv5 zimi kanje:
1- Faka i-PyTorch: Ngenxa yokuthi i-YOLOv5 isekelwe kuhlaka lwe-PyTorch, kufanele uqale ufake i-PyTorch.
pip install torch torchvision
2. Faka i-CUDA: Kumelwe ufake i-CUDA uma uhlose ukusebenzisa i-YOLOv5 ku-GPU.
3. Faka i-YOLOv5: Ngemva kokusetha i-PyTorch ne-CUDA, sebenzisa umyalo olandelayo ukuze ulande i-YOLOv5.
pip install yolov5
4-Ngokulandela ukufakwa kwe-YOLOv5, kufanele ulande izisindo eziqeqeshwe ngaphambilini. Izisindo eziqeqeshwe ngaphambilini ziyatholakala ku-Ultralytics GitHub repo.
Iya engxenyeni "yesisindo" yewebhusayithi ngokuskrolela phansi. Ungalanda izisindo eziqeqeshwe ngaphambilini ohlwini ongaluthola lapha.
5. Khetha izisindo eseziqeqeshiwe futhi ezifanelana kangcono necala lakho lokusebenzisa. Idathasethi noma inguqulo ethile ye-YOLOv5 okufundwe izisindo ingase isetshenziselwe ukunciphisa uhlu.
6- Ngemva kokukhetha izisindo ezifanele, khetha isisindo ngokuchofoza inkinobho ethi “Landa” eduze kwayo. Izisindo zizotholakala ukuze zilandwe njenge. pt amafayela.
7- Dlulisela izisindo ezilandiwe kusiqondisi. Lapha yilapho umbhalo wakho wokuthola uzobe usebenza khona.
8- Kuleli qophelo, ungaqalisa ukutholwa kwento ezithombeni noma amavidiyo akho usebenzisa izisindo eziqeqeshwe kusengaphambili kusikripthi sakho sokuthola.
Lungiselela Idatha
Kufanele wenze lezi zinyathelo ezilandelayo ukuze ulungise idatha ukuze isetshenziswe nge-YOLOv5:
1. Qoqa idatha: Isinyathelo sokuqala ukuqoqa idatha yesithombe noma yevidiyo ozoyidinga ukutholwa kwento. Izinto ofisa ukuzithola kufanele zibe khona ezithombeni noma kumavidiyo.
2- Fometha idatha: Ungavele ungenise izithombe kuskripthi sakho uma uzisebenzisa. Kufanele uguqule ividiyo ibe uchungechunge lwezithombe uma uhlela ukusisebenzisa esisodwa. Ungakhipha amafreyimu ku-movie usebenzisa umtapo wezincwadi onjenge-OpenCV.
import cv2
img = cv2.imread('path/to/image')
Ngomtapo wezincwadi we-OpenCV, ungasebenzisa umyalo olandelayo ukuze uguqule ividiyo ibe uchungechunge lwezithombe:
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. Faka ilebula kudatha: Kumelwe ulebule idatha uma usebenzisa idathasethi yakho. Ukudweba amabhokisi ahlanganisayo azungeze izinto ofisa ukuzibona kuhlaka ngalunye lwesithombe. Kuyinqubo yokulebula idatha. Ungasebenzisa amathuluzi amaningana ukukusiza ngalo msebenzi, okuhlanganisa i-LabelImg ne-RectLabel.
4- Kumelwe uhlukanise idatha ibe amasethi okuqeqesha nokuhlola ngemva kokuyimaka. Lokhu kubalulekile ekuhloleni ukuthi imodeli yakho isebenza kahle kangakanani.
5. Okokugcina, ungase udinge ukucubungula idatha ngaphambi kokuqeqeshwa noma ukuhlolwa. Lokhu kungabandakanya ukukala izithombe noma amavidiyo, ukwenza amanani amaphikseli afane, noma ukusebenzisa izindlela zokwandisa idatha.
Ngemva kokuqeda lezi zinyathelo, idatha yakho isilungile.
Qalisa iskripthi sokuthola
Nawu umfanekiso weskripthi sokuthola esihlaziya isithombe bese sithola izinto.
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))
Ukusebenza ngemuva
Ukucindezelwa okungeyona ingxenye ephezulu kungenye yezindlela ezivame kakhulu zokucubungula emva kokucubungula ezisetshenziswa ekutholeni into (i-NMS). Sisebenzisa i-NMS ukususa amabhokisi abophayo agqagqene ento efanayo. Ukuze sisebenzise i-NMS ekutholeni, singasebenzisa indlela ye-OpenCV library cv2.dnn.NMSBoxes().
Nasi isibonelo sendlela yokuthola ngemuva kokucubungula usebenzisa i-NMS.
import cv2
# Perform Non-Maxima Suppression (NMS)
indices = cv2.dnn.NMSBoxes(ukutholwa, okuyimfihlo, conf_thres, nms_thres)
Okubona
Endabeni yokubuka ngeso, singaphinda sisebenzise umtapo wezincwadi onjenge-OpenCV. Singabonisa amabhokisi ahlanganisayo azungeze izinto ezitholiwe esithombeni esiwumthombo noma ividiyo. Ukuze udwebe amabhokisi okubopha isithombe, sebenzisa indlela ye-cv2.rectangle(). Nansi indlela yokubuka ukutholwa esithombeni sokuqala:
ngenisa cv2
# Draw the bounding boxes on the image
ngoba mina nge indices:
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
Izicelo
I-YOLOv5 imodeli eqinile yokuhlonza into. Ngakho-ke, singayisebenzisa ezimweni eziningi zomhlaba wangempela. Okunye okugqama kakhulu ukusetshenziswa yizimoto ezizishayelayo. I-YOLOv5 ingahlonza izinto ngesikhathi sangempela njengezimoto namarobhothi.
Kumasistimu okuqapha, singasebenzisa i-YOLOv5 ukuze sibone futhi silandelele izinto ekusakazweni kwamavidiyo abukhoma. Ngaphezu kwalokho, i-YOLOv5 ingaba yimpahla enhle kumarobhothi. Kungasiza amarobhothi ukuthi abone futhi aqonde indawo ewazungezile. Lokhu kubaluleke kakhulu emisebenzini efana nokuzulazula nokukhohlisa.
I-YOLOv5 ingase futhi isetshenziswe kunoma iyiphi imboni edinga ukutholwa kwento, njengokudayisa, ezemidlalo, ezokwelashwa, nokuphepha.
Isiphetho
Okokugcina, i-YOLOv5 inguqulo yakamuva neyinkimbinkimbi kakhulu yomndeni we-YOLO we ukutholwa kwento Amamodeli
. Futhi, kufanelekile ukusho ukuthi iyimodeli enembile kakhulu yokuthola into etholakalayo. Ngenxa yokunemba kwayo okuphezulu nesivinini, ungayikhetha ngokuphephile kumaphrojekthi akho okuthola into.
Resky Agus
Ngenza ijenali yokuqala mayelana nemoto yokuthola nge-yolov5 kanye nale webhu isiza ukusesha ulwazi ngalokho.
Ngithanda kakhulu i-AI.
uma ukwazi nginombuzo omningi nge-AI mhlawumbe ungangisiza
Ngiyabonga