Tekinoloje yozindikira zinthu zamakompyuta ndiyofunikira pamapulogalamu ambiri. Timagwiritsa ntchito ma robotiki, zida zowunikira, magalimoto odziyendetsa okha, ndi zina zambiri. Chifukwa chake, timapeza ndikuzindikira zinthu zina mu chithunzi kapena kanema.
Mmodzi mwa odziwika kwambiri ma algorithms ozindikiritsa chinthu ndi YOLO (Mumangoyang'ana Kamodzi) mndandanda wamitundu. Zitsanzozi zimapangidwa ndi Malingaliro a kampani Ultralytics LLC
Mtundu waposachedwa kwambiri wa mndandandawu ndi YOLOv5. Ndipo, ndiye njira yofulumira komanso yolondola kwambiri yozindikiritsira chinthu pamsika. Kuthekera kwachitsanzo kusinthira ku data yatsopano kwasinthidwa kwambiri. Komanso, ili ndi zinthu zambiri zomwe zimapangitsa kuti izichita bwino kuposa momwe zimakhalira kale.
YOLOv5 ndiyabwino kugwiritsa ntchito nthawi yeniyeni chifukwa imatha kukonza zithunzi pamlingo wofikira mafelemu 1000 pa sekondi imodzi pa GPU imodzi.
M'nkhaniyi, tikuwonetsani YOLOv5 ndikufotokozera tsatanetsatane wa madera ake.
Ulendo wa YOLO: Kuchokera ku YOLO kupita ku YOLOv5
Joseph Redmon et al. poyambilira adayambitsa YOLO, mndandanda wazinthu zozindikiritsa zinthu, mu 2016. Mtundu woyamba wa YOLO ukhoza kuzindikira zinthu munthawi yeniyeni. Komabe, inali yolondola pang'ono poyerekeza ndi zitsanzo zina panthawiyo.
Mitundu ingapo yosinthidwa ya YOLO idatulutsidwa zaka zonse. Ndipo pomaliza, Ultralytics LLC idapanga mtundu watsopano wa YOLO, YOLOv5.
YOLOv5 ndiye njira yolondola kwambiri komanso yachangu kwambiri yozindikiritsira chinthu yomwe ilipo pakadali pano.
Zofunika Kwambiri
Mabokosi a Anchor
YOLOv5 imalosera mabokosi omangirira azinthu pachithunzi pogwiritsa ntchito mabokosi a nangula. Chitsanzocho chimalosera kuti ndi ndani mwa mabokosi omwe atchulidwa kale omwe ali ndi magawo osiyanasiyana omwe amafanana bwino ndi chinthu chomwe chili pachithunzichi pogwiritsa ntchito mabokosi a nangula. Awa ndi mabokosi ofotokozedwatu.
Ndipo, amalola YOLOv5 kuzindikira ndi kupeza zinthu pa chithunzi molondola.
Kuchulukitsa kwa data ya Mose
Pophunzitsa, YOLOv5 amagwiritsa ntchito njira yotchedwa mosaic kuwonjezera deta. Kupanga zithunzi zatsopano zophunzitsira, chitsanzo chathu chimaphatikiza zigamba za zithunzi zingapo mwachisawawa. Zotsatira zake, chitsanzocho chimakhala chokhazikika komanso chodalirika. Chifukwa chake, imafikira kuzinthu zatsopano ndikuchepetsa kuchulukitsa.
Chitoliro Chapadera Chophunzitsira
Paipi yophunzitsira yapadera yomwe imasakanikirana kuyang'aniridwa ndi maphunziro osayang'aniridwa amagwiritsidwa ntchito.
Chifukwa chake, chitsanzocho chimaphunzira kuchokera kuchitsanzo chaching'ono ndipo chimagwiritsa ntchito mawu osalembedwa bwino. Izi zimakulitsa magwiridwe antchito achitsanzocho ndikuwonjezera kuthekera kwake kophatikiza zatsopano.
Zigawo zomwe zatsalira komanso zosatsalira
Zomangamanga za YOLOv5 zimaphatikiza zigawo zomwe ndi zotsalira komanso zosatsalira. Polola kuti ma gradients aziyenda m'magawo, zigawo zotsalira zimathandizira chitsanzo kuphunzira zinthu zovuta. Komanso, zigawo zosatsalira zimapatsa chitsanzocho kumvetsetsa bwino kwa chithunzi cholowetsa. Zotsatira zake, YOLOv5 imatha kugwira ntchito moyenera komanso moyenera.
Momwe Mungagwiritsire Ntchito YOLOv5
unsembe
Kuyika kwa YOLOv5 kumatha kumalizidwa mwachangu pogwiritsa ntchito pip. Pip ndi woyang'anira phukusi la Python. Njira zambiri zoyika YOLOv5 ndi izi:
1- Ikani PyTorch: Chifukwa YOLOv5 imachokera ku PyTorch chimango, muyenera choyamba kukhazikitsa PyTorch.
pip install torch torchvision
2. Ikani CUDA: Muyenera kukhazikitsa CUDA ngati mukufuna kuyendetsa YOLOv5 pa GPU.
3. Ikani YOLOv5: Mutatha kukhazikitsa PyTorch ndi CUDA, gwiritsani ntchito lamulo ili kuti mutsitse YOLOv5.
pip install yolov5
4-Kutsatira kuyika kwa YOLOv5, muyenera kutsitsa masikelo omwe mwaphunzitsidwa kale. Zolemera zophunzitsidwa kale zimapezeka mu Ultralytics GitHub repo.
Pitani ku gawo la "zolemera" la webusayiti podutsa pansi. Mutha kutsitsa zolemetsa zophunzitsidwa kale pamndandanda womwe mungapeze apa.
5. Sankhani zolemera zomwe zaphunzitsidwa kale ndipo zigwirizane ndi momwe mungagwiritsire ntchito. Deta kapena mtundu wina wa YOLOv5 womwe masikelo adaphunziridwa angagwiritsidwe ntchito kuchepetsa mndandanda.
6- Mukasankha zolemera zoyenera, sankhani kulemera kwake podina batani la "Koperani" pafupi ndi izo. Zolemerazo zitha kutsitsa ngati. pt mafayilo.
7- Tumizani zolemera zomwe zidatsitsidwa ku chikwatu. Apa ndipamene script yanu yodziwira idzagwira ntchito.
8- Pakadali pano, mutha kuyendetsa kuzindikira kwa zinthu pazithunzi kapena makanema anu pogwiritsa ntchito masikelo ophunzitsidwa kale muzolemba zanu.
Konzani Deta
Muyenera kuchita izi kuti mukonzekere kugwiritsidwa ntchito ndi YOLOv5:
1. Sonkhanitsani deta: Chinthu choyamba ndikusonkhanitsa chithunzi kapena kanema deta yomwe mungafunikire kuzindikira kwa chinthu. Zinthu zomwe mukufuna kuzizindikira ziyenera kupezeka muzithunzi kapena makanema.
2- Sinthani zambiri: Mutha kungolowetsa zithunzi muzolemba zanu ngati mukuzigwiritsa ntchito. Muyenera kusandutsa kanema kukhala zithunzi zingapo ngati mukufuna kugwiritsa ntchito imodzi. Mutha kuchotsa mafelemu mu kanema pogwiritsa ntchito laibulale ngati OpenCV.
import cv2
img = cv2.imread('path/to/image')
Ndi laibulale ya OpenCV, mutha kugwiritsa ntchito lamulo ili kuti musinthe kanema kukhala zithunzi zingapo:
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. Lembani deta: Muyenera kulemba deta ngati mukugwiritsa ntchito deta yanu. Kujambula mabokosi omangira mozungulira zinthu zomwe mukufuna kuzizindikira muzithunzi zilizonse. Ndi ndondomeko yolembera deta. Mutha kugwiritsa ntchito zida zingapo kukuthandizani ndi ntchitoyi, kuphatikiza LabelImg ndi RectLabel.
4- Muyenera kugawanitsa deta muzophunzitsira ndi zoyeserera mutazilemba. Izi ndizofunikira kuti muwone momwe chitsanzo chanu chikuyendera bwino.
5. Pomaliza, mungafunike kukonzetseratu deta musanaphunzire kapena kuyezetsa. Izi zitha kuphatikizapo kukulitsa zithunzi kapena makanema, kuyika miyeso ya pixel, kapena kugwiritsa ntchito njira zowonjezera deta.
Mukamaliza masitepe awa, deta yanu ndi yokonzeka.
Yambitsani script yozindikira
Nachi fanizo la script yozindikira yomwe imasanthula chithunzi ndikupeza zinthu.
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))
Kutumiza pambuyo
Kuponderezedwa kosalekeza ndi imodzi mwa njira zomwe zimagwiritsidwa ntchito pofufuza zinthu (NMS). Timagwiritsa ntchito NMS kuchotsa mabokosi omangirira a chinthu chomwecho. Kuti tigwiritse ntchito NMS pazozindikira, titha kugwiritsa ntchito njira ya laibulale ya OpenCV cv2.dnn.NMSBoxes().
Nachi chitsanzo cha momwe mungasinthire zozindikira pogwiritsa ntchito NMS.
import cv2
# Perform Non-Maxima Suppression (NMS)
indices = cv2.dnn.NMSBoxes(zozindikira, zinsinsi, conf_thres, nms_thres)
Kuwonetseratu
Pankhani yowonera, titha kugwiritsanso ntchito laibulale ngati OpenCV. Titha kuwonetsa mabokosi omangika mozungulira zinthu zomwe zapezeka pazithunzi kapena kanema. Kuti mujambule mabokosi omangira chithunzi, gwiritsani ntchito njira ya cv2.rectangle(). Umu ndi momwe mungawonere zomwe zazindikirika pachithunzi choyambirira:
tumizani cv2
# Draw the bounding boxes on the image
pakuti ine mu 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
Mapulogalamu
YOLOv5 ndi chitsanzo champhamvu chozindikiritsa chinthu. Chifukwa chake, titha kugwiritsa ntchito zochitika zenizeni padziko lapansi. Chimodzi mwazogwiritsidwa ntchito kwambiri ndi magalimoto odziyendetsa okha. YOLOv5 imatha kuzindikira zinthu munthawi yeniyeni monga magalimoto ndi magetsi apamsewu.
M'machitidwe owunikira, titha kugwiritsa ntchito YOLOv5 kuzindikira ndikutsata zinthu zomwe zili m'mavidiyo amoyo. Kuphatikiza apo, YOLOv5 ikhoza kukhala yothandiza kwambiri pama robotiki. Itha kuthandiza maloboti kuzindikira ndikumvetsetsa komwe amakhala. Izi ndi zofunika kwambiri pazochitika monga kuyenda ndi kusintha.
YOLOv5 itha kugwiritsidwanso ntchito pamakampani aliwonse omwe amafunikira kuzindikira zinthu, monga malonda, masewera, zamankhwala, ndi chitetezo.
Kutsiliza
Pomaliza, YOLOv5 ndiye mtundu waposachedwa komanso wovuta kwambiri wa banja la YOLO la kuzindikira kwa chinthu zitsanzo
. Komanso, nkoyenera kunena kuti ndi njira yolondola kwambiri yodziwira zinthu yomwe ilipo. Chifukwa cha kulondola kwake komanso kuthamanga kwake, mutha kuyisankha mosamala pama projekiti anu ozindikira zinthu.
Resky Agus
Ndimapanga magazini yoyamba yokhudza galimoto yodziwika ndi yolov5 ndi chithandizo cha intaneti ichi kuti ndifufuze zambiri za izo.
Ndimakonda kwambiri AI.
ngati mungathe ndili ndi mafunso ambiri okhudza AI mwina mutha kundithandiza
Zikomo