Ithala leencwadi lasimahla i-OpenCV 4.7 (iThala leeNcwadi loMbono weKhompyutha enguMthombo oVulekileyo) lakhululwa, libonelela ngezixhobo zokusebenza kunye nokuhlalutya umxholo wemifanekiso. I-OpenCV ibonelela ngee-algorithms ezingaphezu kwama-2500, zombini zakudala kwaye zibonisa inkqubela phambili yamva nje kumbono wekhompyuter kunye neenkqubo zokufunda koomatshini. Ikhowudi yelayibrari ibhalwe kwi-C ++ kwaye isasazwe phantsi kwelayisensi ye-BSD. Izibophelelo zilungiselelwe iilwimi ezahlukeneyo zokucwangcisa, kuquka iPython, iMATLAB kunye neJava.
Ithala leencwadi lingasetyenziselwa ukuqaphela izinto kwiifoto kunye neevidiyo (umzekelo, ukuqatshelwa kobuso kunye namanani abantu, umbhalo, njl.), ukulandelela intshukumo yezinto kunye neekhamera, ukuhlelwa kweentshukumo kwividiyo, ukuguqula imifanekiso, ukukhupha iimodeli ze-3D, ukuvelisa indawo ye-3D evela kwimifanekiso esuka kwiikhamera ze-stereo, ukudala imifanekiso ekumgangatho ophezulu ngokudibanisa imifanekiso ephantsi, ukukhangela izinto ezikumfanekiso ezifana nesethi yezinto eziveziweyo, ukusebenzisa iindlela zokufunda umatshini, ukubeka iimpawu, ukuchonga izinto eziqhelekileyo kwizinto ezahlukeneyo. imifanekiso, isusa ngokuzenzekelayo iziphene ezifana neliso elibomvu .
Phakathi kotshintsho kukhupho olutsha:
- Ukulungiswa okubalulekileyo kokusebenza kwe-convolution kwimodyuli ye-DNN (i-Deep Neural Network) iqhutywe kunye nokuphunyezwa kwe-algorithms yokufunda koomatshini ngokusekelwe kwiinethiwekhi ze-neural. I-algorithm ye-Winograd ye-convolution ngokukhawuleza iphunyeziwe. Kongezwe iileyile ezintsha ze-ONNX (Vula iNeural Network Exchange): I-Scatter, ScatterND, Tile, ReduceL1 kunye ne-ReduceMin. Inkxaso eyongeziweyo ye-OpenVino 2022.1 isakhelo kunye ne-CANN backend.
- Ubulunga obuphuculweyo bokufunyanwa kwekhowudi ye-QR kunye nokuchazwa kwekhowudi.
- Inkxaso eyongeziweyo yabamakishi ababonakalayo i-ArUco kunye ne-AprilTag.
- Yongezwe i-Nanotrack v2 tracker esekwe kwinethiwekhi ye-neural.
- Kuphunyezwe i-Stackblur blur algorithm.
- Inkxaso eyongeziweyo yeFFmpeg 5.x kunye neCUDA 12.0.
- I-API entsha iye yacetywa ukuze ilawule iifomati zemifanekiso enamaphepha amaninzi.
- Inkxaso eyongeziweyo yelayibrari ye-libSPNG yefomathi yePNG.
- i-libJPEG-Turbo yenza unikezelo usebenzisa imiyalelo ye-SIMD.
- Kwiqonga le-Android, inkxaso ye-H264 / H265 iphunyeziwe.
- Zonke ezisisiseko Python APIs zinikezelwe.
- Kongezwe i-backend entsha yendalo yonke yemiyalelo ye-vector.
umthombo: opennet.ru