{"id":39125,"date":"2026-02-15T01:00:42","date_gmt":"2026-02-14T22:00:42","guid":{"rendered":"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/"},"modified":"2026-02-15T01:00:42","modified_gmt":"2026-02-14T22:00:42","slug":"haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi","status":"publish","type":"post","link":"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/","title":{"rendered":"Haar Cascades’ten YOLO’ya: Y\u00fcz Alg\u0131lama Sistemlerini Modernize Etme Rehberi"},"content":{"rendered":"

Haar Cascades’ten YOLO’ya: Y\u00fcz Alg\u0131lama Sistemlerini Modernize Etme Rehberi<\/h2>\n

Bilgisayar g\u00f6r\u00fc\u015f\u00fc alan\u0131nda y\u00fcz alg\u0131lama, y\u0131llard\u0131r temel bir ara\u015ft\u0131rma ve uygulama konusu olmu\u015ftur. Geleneksel y\u00f6ntemlerden modern derin \u00f6\u011frenme yakla\u015f\u0131mlar\u0131na uzanan bu yolculukta, Haar Cascades gibi klasik algoritmalar yerini YOLO (You Only Look Once) gibi ger\u00e7ek zamanl\u0131, y\u00fcksek performansl\u0131 modellere b\u0131rakmaktad\u0131r. Bu rehber, mevcut Haar Cascades tabanl\u0131 y\u00fcz alg\u0131lama sistemlerini, derin \u00f6\u011frenmenin g\u00fcc\u00fcn\u00fc kullanan YOLO modellerine nas\u0131l ta\u015f\u0131yaca\u011f\u0131n\u0131z\u0131 ad\u0131m ad\u0131m a\u00e7\u0131klamaktad\u0131r. Amac\u0131m\u0131z, daha h\u0131zl\u0131, daha do\u011fru ve daha g\u00fcvenilir y\u00fcz alg\u0131lama \u00e7\u00f6z\u00fcmlerine ge\u00e7i\u015finizi kolayla\u015ft\u0131rmakt\u0131r.<\/p>\n

Y\u00fcz Alg\u0131laman\u0131n Evrimi: Haar Cascades’ten Derin \u00d6\u011frenmeye<\/h3>\n

Y\u00fcz alg\u0131lama teknolojisi, ilk basit \u00f6zellik tabanl\u0131 yakla\u015f\u0131mlardan g\u00fcn\u00fcm\u00fcz\u00fcn karma\u015f\u0131k sinir a\u011f\u0131 modellerine kadar uzun bir yol kat etmi\u015ftir. Bu evrim, hem alg\u0131lama do\u011frulu\u011funu hem de h\u0131z\u0131n\u0131 \u00f6nemli \u00f6l\u00e7\u00fcde art\u0131rm\u0131\u015ft\u0131r.<\/p>\n

Haar Cascades: Temelleri ve \u00c7al\u0131\u015fma Prensibi<\/h4>\n

Haar Cascades, Paul Viola ve Michael Jones taraf\u0131ndan 2001 y\u0131l\u0131nda tan\u0131t\u0131lan, makine \u00f6\u011frenimi tabanl\u0131 bir nesne alg\u0131lama \u00e7er\u00e7evesidir. \u00d6zellikle y\u00fcz alg\u0131lama i\u00e7in pop\u00fcler hale gelmi\u015ftir. Temelinde, bir g\u00f6r\u00fcnt\u00fcn\u00fcn farkl\u0131 b\u00f6lgelerindeki piksel yo\u011funluk farklar\u0131n\u0131 \u00f6l\u00e7en “Haar benzeri \u00f6zellikler” bulunur. Bu \u00f6zellikler, bir y\u00fcz\u00fcn g\u00f6zleri, burnu veya a\u011fz\u0131 gibi belirli yap\u0131lar\u0131n\u0131 temsil edebilir. Alg\u0131lama s\u00fcreci, AdaBoost algoritmas\u0131 ile e\u011fitilmi\u015f bir dizi “g\u00fc\u00e7l\u00fc s\u0131n\u0131fland\u0131r\u0131c\u0131” (weak classifiers) kullan\u0131r. Bu s\u0131n\u0131fland\u0131r\u0131c\u0131lar, bir kaskad yap\u0131s\u0131nda d\u00fczenlenir; her a\u015fama, bir \u00f6nceki a\u015famay\u0131 ge\u00e7en aday b\u00f6lgeleri daha detayl\u0131 inceler. Bu sayede, y\u00fcz olmayan b\u00f6lgeler erken a\u015famalarda elenerek hesaplama y\u00fck\u00fc azalt\u0131l\u0131r.<\/p>\n

Haar Cascades’in Avantajlar\u0131 ve S\u0131n\u0131rl\u0131l\u0131klar\u0131<\/h4>\n

Haar Cascades’in en b\u00fcy\u00fck avantaj\u0131, nispeten d\u00fc\u015f\u00fck hesaplama g\u00fcc\u00fc gerektirmesi ve bu nedenle eski donan\u0131mlarda bile \u00e7al\u0131\u015fabilmesidir. Ayr\u0131ca, OpenCV gibi k\u00fct\u00fcphanelerde haz\u0131r e\u011fitimli modellerin bulunmas\u0131, h\u0131zl\u0131 prototipleme imkan\u0131 sunar. Ancak, \u00f6nemli s\u0131n\u0131rl\u0131l\u0131klar\u0131 da vard\u0131r:<\/p>\n

* Do\u011fruluk:<\/strong> \u00d6zellikle farkl\u0131 ayd\u0131nlatma ko\u015fullar\u0131, y\u00fcz pozisyonlar\u0131 (a\u00e7\u0131, e\u011fim) ve k\u0131smi kapanmalar kar\u015f\u0131s\u0131nda performans\u0131 d\u00fc\u015febilir.
\n* Yanl\u0131\u015f Pozitifler:<\/strong> Bazen y\u00fcz olmayan b\u00f6lgeleri y\u00fcz olarak yanl\u0131\u015fl\u0131kla alg\u0131layabilir.
\n* \u00d6l\u00e7eklenebilirlik:<\/strong> Tek bir nesne s\u0131n\u0131f\u0131 i\u00e7in e\u011fitilir ve yeni nesneler i\u00e7in yeniden e\u011fitim s\u00fcreci karma\u015f\u0131kt\u0131r.
\n* Ger\u00e7ek Zamanl\u0131 Performans:<\/strong> Y\u00fcksek \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc g\u00f6r\u00fcnt\u00fclerde veya \u00e7ok say\u0131da y\u00fcz i\u00e7eren sahnelerde ger\u00e7ek zamanl\u0131 performans\u0131 yetersiz kalabilir.<\/p>\n

Derin \u00d6\u011frenmenin Y\u00fckseli\u015fi ve Bilgisayar G\u00f6r\u00fcs\u00fcne Etkisi<\/h4>\n

2010’lu y\u0131llar\u0131n ba\u015f\u0131ndan itibaren derin \u00f6\u011frenme, bilgisayar g\u00f6r\u00fc\u015f\u00fc alan\u0131nda devrim niteli\u011finde de\u011fi\u015fiklikler yaratm\u0131\u015ft\u0131r. Evri\u015fimli Sinir A\u011flar\u0131 (CNN’ler), g\u00f6r\u00fcnt\u00fc s\u0131n\u0131fland\u0131rma, nesne alg\u0131lama ve segmentasyon gibi g\u00f6revlerde insan\u00fcst\u00fc performans sergilemeye ba\u015flam\u0131\u015ft\u0131r. B\u00fcy\u00fck veri setleri ve geli\u015fmi\u015f donan\u0131mlar (GPU’lar) sayesinde, derin \u00f6\u011frenme modelleri, geleneksel y\u00f6ntemlerin \u00f6tesine ge\u00e7erek daha karma\u015f\u0131k ve soyut \u00f6zellikleri otomatik olarak \u00f6\u011frenebilme yetene\u011fi kazanm\u0131\u015ft\u0131r. Bu durum, y\u00fcz alg\u0131lama sistemlerinin do\u011frulu\u011funu ve sa\u011flaml\u0131\u011f\u0131n\u0131 \u00f6nemli \u00f6l\u00e7\u00fcde art\u0131rm\u0131\u015ft\u0131r.<\/p>\n

YOLO Nedir ve Neden Y\u00fcz Alg\u0131lama \u0130\u00e7in Tercih Edilmeli?<\/h3>\n

Derin \u00f6\u011frenmenin nesne alg\u0131lama alan\u0131ndaki en \u00f6nemli ba\u015far\u0131lar\u0131ndan biri, YOLO (You Only Look Once) mimarisidir. YOLO, hem h\u0131z\u0131 hem de do\u011frulu\u011fu bir araya getirerek ger\u00e7ek zamanl\u0131 uygulamalar i\u00e7in ideal bir \u00e7\u00f6z\u00fcm sunar.<\/p>\n

YOLO’nun Mimarisi ve Ger\u00e7ek Zamanl\u0131 Alg\u0131lama<\/h4>\n

YOLO, g\u00f6r\u00fcnt\u00fclerdeki nesneleri tek bir sinir a\u011f\u0131 ge\u00e7i\u015fiyle alg\u0131layan devrim niteli\u011finde bir yakla\u015f\u0131md\u0131r. Geleneksel nesne alg\u0131lama y\u00f6ntemleri (\u00f6rne\u011fin R-CNN ailesi), \u00f6nce potansiyel nesne b\u00f6lgelerini belirler, ard\u0131ndan bu b\u00f6lgeleri s\u0131n\u0131fland\u0131r\u0131r. YOLO ise g\u00f6r\u00fcnt\u00fcy\u00fc do\u011frudan bir \u0131zgaraya b\u00f6ler ve her \u0131zgara h\u00fccresinin belirli say\u0131da s\u0131n\u0131rlay\u0131c\u0131 kutu (bounding box) ve bu kutular\u0131n nesne i\u00e7erme olas\u0131l\u0131klar\u0131n\u0131 tahmin etmesini sa\u011flar. Ayn\u0131 zamanda, her kutu i\u00e7in s\u0131n\u0131f olas\u0131l\u0131klar\u0131n\u0131 da tahmin eder. Bu “tek seferde bakma” prensibi, YOLO’yu inan\u0131lmaz derecede h\u0131zl\u0131 hale getirir ve saniyede kare (FPS) cinsinden y\u00fcksek performans sunar.<\/p>\n

YOLO’nun Y\u00fcz Alg\u0131lamadaki \u00dcst\u00fcnl\u00fckleri<\/h4>\n

YOLO’nun y\u00fcz alg\u0131lama i\u00e7in tercih edilmesinin bir\u00e7ok nedeni vard\u0131r:<\/p>\n

* Ger\u00e7ek Zamanl\u0131 Performans:<\/strong> Kamera ak\u0131\u015flar\u0131 veya video analizleri gibi ger\u00e7ek zamanl\u0131 uygulamalar i\u00e7in kritik olan y\u00fcksek FPS de\u011ferlerine ula\u015fabilir.
\n* Y\u00fcksek Do\u011fruluk:<\/strong> Geni\u015f veri setleri \u00fczerinde e\u011fitildi\u011finde, farkl\u0131 pozisyon, ayd\u0131nlatma ve \u00f6l\u00e7eklerdeki y\u00fczleri y\u00fcksek do\u011frulukla alg\u0131layabilir.
\n* Sa\u011flaml\u0131k:<\/strong> K\u0131smi kapanmalar, farkl\u0131 y\u00fcz ifadeleri ve kalabal\u0131k sahneler gibi zorlu ko\u015fullarda Haar Cascades’e g\u00f6re \u00e7ok daha iyi performans g\u00f6sterir.
\n* Tek Model:<\/strong> Tek bir modelle birden fazla y\u00fcz\u00fc ayn\u0131 anda ve verimli bir \u015fekilde alg\u0131layabilir.<\/p>\n

Farkl\u0131 YOLO Versiyonlar\u0131 (YOLOv3, YOLOv4, YOLOv5, YOLOv8)<\/h4>\n

YOLO mimarisi, ilk tan\u0131t\u0131m\u0131ndan bu yana s\u00fcrekli olarak geli\u015ftirilmi\u015ftir. Her yeni versiyon, \u00f6nceki versiyonlar\u0131n eksiklerini gidererek daha iyi performans sunar:<\/p>\n

* YOLOv3:<\/strong> Daha derin bir a\u011f mimarisi (Darknet-53) ve \u00e7oklu \u00f6l\u00e7ekte alg\u0131lama yetene\u011fi ile daha k\u00fc\u00e7\u00fck nesnelerin alg\u0131lanmas\u0131nda iyile\u015fme sa\u011flam\u0131\u015ft\u0131r.
\n* YOLOv4:<\/strong> CSPDarknet53, SPP blo\u011fu, PANet gibi yenilik\u00e7i tekniklerle hem h\u0131z\u0131 hem de do\u011frulu\u011fu \u00f6nemli \u00f6l\u00e7\u00fcde art\u0131rm\u0131\u015ft\u0131r.
\n* YOLOv5:<\/strong> PyTorch tabanl\u0131 ilk resmi YOLO versiyonudur. Daha mod\u00fcler bir yap\u0131, farkl\u0131 boyutlarda modeller (nano, small, medium, large, xlarge) ve kolay kullan\u0131m sunar. Genellikle h\u0131z ve do\u011fruluk aras\u0131nda iyi bir denge sa\u011flar.
\n* YOLOv8:<\/strong> Ultralytics taraf\u0131ndan geli\u015ftirilen en yeni versiyondur. Daha optimize edilmi\u015f bir mimari, geli\u015fmi\u015f e\u011fitim stratejileri ve daha iyi performans sunar. Nesne alg\u0131laman\u0131n yan\u0131 s\u0131ra segmentasyon ve poz tahmini gibi g\u00f6revleri de destekler. Y\u00fcz alg\u0131lama i\u00e7in genellikle en iyi se\u00e7enektir.<\/p>\n

Haar Cascades Tabanl\u0131 Bir Sistemi Anlamak ve De\u011ferlendirmek<\/h3>\n

YOLO’ya ge\u00e7i\u015f yapmadan \u00f6nce mevcut Haar Cascades tabanl\u0131 sisteminizi iyi anlamak ve potansiyel darbo\u011fazlar\u0131n\u0131 belirlemek \u00f6nemlidir.<\/p>\n

Mevcut Haar Cascade Uygulamas\u0131n\u0131 \u0130nceleme<\/h4>\n

\u0130lk ad\u0131m, mevcut Haar Cascade kod taban\u0131n\u0131z\u0131 g\u00f6zden ge\u00e7irmektir. Genellikle Python ve OpenCV kullan\u0131larak yaz\u0131lm\u0131\u015f bu uygulamalarda cv2.CascadeClassifier<\/code> s\u0131n\u0131f\u0131 ve detectMultiScale<\/code> metodu merkezi rol oynar. Bu metot, g\u00f6r\u00fcnt\u00fcy\u00fc farkl\u0131 \u00f6l\u00e7eklerde tarayarak y\u00fczleri alg\u0131lar. \u0130nceleme s\u0131ras\u0131nda, kullan\u0131lan Haar cascade XML dosyas\u0131n\u0131n (\u00f6rne\u011fin haarcascade_frontalface_default.xml<\/code>) hangi y\u00fcz tipleri i\u00e7in e\u011fitildi\u011fini, scaleFactor<\/code>, minNeighbors<\/code>, minSize<\/code> gibi parametrelerin de\u011ferlerini ve bunlar\u0131n alg\u0131lama \u00fczerindeki etkilerini anlamaya \u00e7al\u0131\u015f\u0131n.<\/p>\n

\nimport cv2\n\n# Haar Cascade y\u00fckle\nface_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')\n\n# G\u00f6r\u00fcnt\u00fcy\u00fc y\u00fckle\nimg = cv2.imread('ornek_yuz.jpg')\ngray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)\n\n# Y\u00fczleri alg\u0131la\nfaces = face_cascade.detectMultiScale(gray, scaleFactor=1.1, minNeighbors=5, minSize=(30, 30))\n\n# Alg\u0131lanan y\u00fczleri \u00e7iz\nfor (x, y, w, h) in faces:\n    cv2.rectangle(img, (x, y), (x+w, y+h), (255, 0, 0), 2)\n\ncv2.imshow('Haar Cascade Yuz Algilama', img)\ncv2.waitKey(0)\ncv2.destroyAllWindows()\n<\/pre>\n
<\/code><\/p>\n
Performans Metrikleri ve Darbo\u011fazlar<\/h4>\n
Mevcut sisteminizin performans\u0131n\u0131 objektif olarak de\u011ferlendirmek i\u00e7in belirli metrikler kullanmal\u0131s\u0131n\u0131z:<\/p>\n
*   FPS (Frames Per Second):<\/strong> Sistemin saniyede ka\u00e7 kare i\u015fleyebildi\u011fini \u00f6l\u00e7er. Ger\u00e7ek zamanl\u0131 uygulamalar i\u00e7in kritik bir metrik.
\n*   Do\u011fruluk (Accuracy):<\/strong> Do\u011fru alg\u0131lanan y\u00fczlerin toplam y\u00fcz say\u0131s\u0131na oran\u0131.
\n*   Yanl\u0131\u015f Pozitifler (False Positives):<\/strong> Y\u00fcz olmayan b\u00f6lgelerin y\u00fcz olarak alg\u0131lanmas\u0131.
\n*   Yanl\u0131\u015f Negatifler (False Negatifler):<\/strong> Ger\u00e7ek y\u00fczlerin alg\u0131lanamamas\u0131.
\n*   Gecikme (Latency):<\/strong> Bir g\u00f6r\u00fcnt\u00fcn\u00fcn i\u015flenmesi i\u00e7in ge\u00e7en s\u00fcre.<\/p>\n
Bu metrikleri farkl\u0131 ayd\u0131nlatma ko\u015fullar\u0131, y\u00fcz pozisyonlar\u0131 ve kalabal\u0131k sahneler gibi \u00e7e\u015fitli senaryolarda \u00f6l\u00e7erek Haar Cascades'in zay\u0131f noktalar\u0131n\u0131 ve YOLO'ya ge\u00e7i\u015fin potansiyel faydalar\u0131n\u0131 daha net g\u00f6rebilirsiniz. Genellikle Haar Cascades'in darbo\u011fazlar\u0131, d\u00fc\u015f\u00fck \u0131\u015f\u0131kta veya yan profildeki y\u00fczleri alg\u0131layamamas\u0131, y\u00fcksek yanl\u0131\u015f pozitif oran\u0131 ve y\u00fcksek \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc videolarda d\u00fc\u015fen FPS de\u011ferleridir.<\/p>\n
Ge\u00e7i\u015f \u0130htiyac\u0131n\u0131 Belirleme<\/h4>\n
Yukar\u0131daki de\u011ferlendirmeler sonucunda, sisteminizin a\u015fa\u011f\u0131daki ihtiya\u00e7lardan bir veya daha fazlas\u0131n\u0131 kar\u015f\u0131layamad\u0131\u011f\u0131n\u0131 fark ederseniz, YOLO'ya ge\u00e7i\u015f yapmak mant\u0131kl\u0131 olacakt\u0131r:<\/p>\n
*   Daha y\u00fcksek alg\u0131lama do\u011frulu\u011fu
\n*   Daha iyi ger\u00e7ek zamanl\u0131 performans
\n*   Farkl\u0131 ko\u015fullara (ayd\u0131nlatma, pozisyon) kar\u015f\u0131 daha fazla sa\u011flaml\u0131k
\n*   Daha az yanl\u0131\u015f pozitif
\n*   Gelecekteki geni\u015flemeler i\u00e7in daha modern ve esnek bir mimari<\/p>\n
YOLO Tabanl\u0131 Y\u00fcz Alg\u0131lamaya Ge\u00e7i\u015f Ad\u0131mlar\u0131<\/h3>\n
Haar Cascades'ten YOLO'ya ge\u00e7i\u015f, yeni bir teknoloji y\u0131\u011f\u0131n\u0131na adapte olmay\u0131 gerektirir. \u0130\u015fte temel ad\u0131mlar:<\/p>\n
Gerekli Ortam Kurulumu ve Ba\u011f\u0131ml\u0131l\u0131klar<\/h4>\n
YOLO modelleri genellikle PyTorch veya TensorFlow gibi derin \u00f6\u011frenme \u00e7er\u00e7eveleri \u00fczerinde \u00e7al\u0131\u015f\u0131r. Ultralytics'in YOLOv5 ve YOLOv8 modelleri genellikle PyTorch tabanl\u0131d\u0131r.<\/p>\n
1.  Python Ortam\u0131:<\/strong> Bir sanal ortam (venv veya Conda) olu\u015fturun.<\/p>\n
\n    conda create -n yolov8_env python=3.9\n    conda activate yolov8_env\n    <\/pre>\n
<\/code>
\n2.  Gerekli K\u00fct\u00fcphaneler:<\/strong> Ultralytics k\u00fct\u00fcphanesini y\u00fckleyin. Bu, gerekli t\u00fcm ba\u011f\u0131ml\u0131l\u0131klar\u0131 (PyTorch, OpenCV vb.) otomatik olarak kuracakt\u0131r.<\/p>\n
\n    pip install ultralytics\n    <\/pre>\n
<\/code>
\n3.  CUDA (GPU Kullan\u0131m\u0131 \u0130\u00e7in):<\/strong> E\u011fer bir NVIDIA GPU'nuz varsa, CUDA ve cuDNN'i kurarak modelin GPU \u00fczerinde \u00e7al\u0131\u015fmas\u0131n\u0131 sa\u011flay\u0131n. Bu, performans\u0131 \u00f6nemli \u00f6l\u00e7\u00fcde art\u0131racakt\u0131r. PyTorch'un resmi belgelerinden do\u011fru CUDA s\u00fcr\u00fcm\u00fcn\u00fc kontrol edin.<\/p>\n
\u00d6nceden E\u011fitilmi\u015f YOLO Modellerini Kullanma<\/h4>\n
YOLO'nun en b\u00fcy\u00fck avantajlar\u0131ndan biri, genellikle geni\u015f veri setleri \u00fczerinde (COCO gibi) \u00f6nceden e\u011fitilmi\u015f modellerin bulunmas\u0131d\u0131r. Y\u00fcz alg\u0131lama i\u00e7in, do\u011frudan y\u00fcz alg\u0131lama \u00fczerine e\u011fitilmi\u015f \u00f6zel YOLO modelleri (\u00f6rne\u011fin, belirli ara\u015ft\u0131rmac\u0131lar taraf\u0131ndan yay\u0131nlanan \"YOLO-Face\" modelleri) veya genel nesne alg\u0131lama modellerinin (YOLOv8n.pt) \"person\" s\u0131n\u0131f\u0131n\u0131 kullanarak insanlar\u0131 alg\u0131lay\u0131p ard\u0131ndan bu b\u00f6lgelerde y\u00fcz arama gibi yakla\u015f\u0131mlar kullan\u0131labilir. Ultralytics'in YOLOv8 modelleri, ultralytics<\/code> k\u00fct\u00fcphanesi ile kolayca y\u00fcklenebilir.<\/p>\n
\nfrom ultralytics import YOLO\n\n# \u00d6nceden e\u011fitilmi\u015f YOLOv8 nano modelini y\u00fckle\n# Not: Bu genel bir nesne alg\u0131lama modelidir.\n# Y\u00fcz alg\u0131lama i\u00e7in \u00f6zel olarak e\u011fitilmi\u015f bir model daha iyi sonu\u00e7lar verecektir.\nmodel = YOLO('yolov8n.pt') \n\n# Veya \u00f6zel bir y\u00fcz alg\u0131lama modeli varsa:\n# model = YOLO('path\/to\/yolov8n-face.pt') \n<\/pre>\n
<\/code><\/p>\n
Veri Haz\u0131rl\u0131\u011f\u0131 ve \u00d6zel Y\u00fcz Veri Setleri (\u0130htiya\u00e7 Halinde)<\/h4>\n
E\u011fer \u00f6nceden e\u011fitilmi\u015f genel modeller mevcut uygulaman\u0131z\u0131n \u00f6zel ihtiya\u00e7lar\u0131n\u0131 kar\u015f\u0131lam\u0131yorsa (\u00f6rne\u011fin, belirli etnik gruplar, maskeli y\u00fczler veya \u00e7ok k\u00fc\u00e7\u00fck y\u00fczler), kendi y\u00fcz veri setinizi olu\u015fturman\u0131z ve modelinizi bu veri \u00fczerinde e\u011fitmeniz gerekebilir.<\/p>\n
1.  Veri Toplama:<\/strong> \u00c7e\u015fitli kaynaklardan (kameralar, internet) y\u00fcz g\u00f6r\u00fcnt\u00fcleri toplay\u0131n. \u00c7e\u015fitlilik (ayd\u0131nlatma, poz, ya\u015f, cinsiyet vb.) \u00e7ok \u00f6nemlidir.
\n2.  Etiketleme (Annotation):<\/strong> Her g\u00f6r\u00fcnt\u00fcdeki y\u00fczleri s\u0131n\u0131rlay\u0131c\u0131 kutularla (bounding boxes) i\u015faretleyin. LabelImg, Roboflow gibi ara\u00e7lar bu s\u00fcre\u00e7te yard\u0131mc\u0131 olabilir. YOLO format\u0131nda etiketler (s\u0131n\u0131f_id x_center y_center width height) olu\u015fturulmal\u0131d\u0131r.
\n3.  Veri B\u00f6lme:<\/strong> Veri setinizi e\u011fitim (train), do\u011frulama (validation) ve test (test) k\u00fcmelerine ay\u0131r\u0131n (genellikle %70\/%20\/%10 oran\u0131 kullan\u0131l\u0131r).<\/p>\n
Model E\u011fitimi ve \u0130nce Ayar (Fine-tuning)<\/h4>\n
Kendi veri setiniz varsa veya mevcut bir modeli \u00f6zel ihtiya\u00e7lar\u0131n\u0131za g\u00f6re uyarlamak istiyorsan\u0131z, model e\u011fitimi veya ince ayar yapman\u0131z gerekir.<\/p>\n
1.  E\u011fitim Konfig\u00fcrasyonu:<\/strong> data.yaml<\/code> dosyas\u0131 olu\u015fturarak veri setinizin yollar\u0131n\u0131 ve s\u0131n\u0131f isimlerini tan\u0131mlay\u0131n.
\n2.  E\u011fitim Komutu:<\/strong> Ultralytics k\u00fct\u00fcphanesi ile e\u011fitim olduk\u00e7a basittir.<\/p>\n
\n    from ultralytics import YOLO\n\n    # \u00d6nceden e\u011fitilmi\u015f bir modeli y\u00fckle (transfer \u00f6\u011frenimi i\u00e7in)\n    model = YOLO('yolov8n.pt') \n\n    # Kendi veri setinizle e\u011fitimi ba\u015flat\n    # 'data.yaml' dosyan\u0131zda veri setinizin yollar\u0131 ve s\u0131n\u0131f bilgileri olmal\u0131\n    # 'epochs' e\u011fitim d\u00f6ng\u00fcs\u00fc say\u0131s\u0131, 'imgsz' giri\u015f g\u00f6r\u00fcnt\u00fcs\u00fc boyutu\n    results = model.train(data='path\/to\/your\/data.yaml', epochs=100, imgsz=640)\n    <\/pre>\n
<\/code>
\n3.  \u0130nce Ayar:<\/strong> Genellikle, b\u00fcy\u00fck bir veri seti \u00fczerinde \u00f6nceden e\u011fitilmi\u015f bir modeli al\u0131p, kendi k\u00fc\u00e7\u00fck veri setiniz \u00fczerinde daha az epoch ile e\u011fitmeye \"ince ayar\" denir. Bu, s\u0131f\u0131rdan e\u011fitime g\u00f6re \u00e7ok daha h\u0131zl\u0131 ve verimlidir.<\/p>\n
Uygulama ve Entegrasyon Pratikleri<\/h3>\n
YOLO modelinizi e\u011fittikten veya haz\u0131r bir modeli y\u00fckledikten sonra, onu uygulaman\u0131za entegre etme zaman\u0131 gelir.<\/p>\n
YOLO Modelini Kullanarak Y\u00fcz Alg\u0131lama (Python \u00d6rne\u011fi)<\/h4>\n
A\u015fa\u011f\u0131da, bir g\u00f6r\u00fcnt\u00fc \u00fczerinde YOLOv8 kullanarak y\u00fcz alg\u0131lama (genel bir modelin 'person' s\u0131n\u0131f\u0131n\u0131 kullanarak) \u00f6rne\u011fi verilmi\u015ftir. Ger\u00e7ek bir y\u00fcz alg\u0131lama i\u00e7in, y\u00fcz alg\u0131lamaya \u00f6zel e\u011fitilmi\u015f bir YOLO modeli tercih edilmelidir.<\/p>\n
\nimport cv2\nfrom ultralytics import YOLO\n\n# YOLOv8 modelini y\u00fckle (\u00f6rne\u011fin, 'yolov8n.pt' genel nesne alg\u0131lama modelidir)\n# Y\u00fcz alg\u0131lama i\u00e7in \u00f6zel olarak e\u011fitilmi\u015f bir model kullanmak daha do\u011fru sonu\u00e7lar verir.\n# E\u011fer b\u00f6yle bir modeliniz varsa: model = YOLO('yolov8n-face.pt')\nmodel = YOLO('yolov8n.pt') \n\n# G\u00f6r\u00fcnt\u00fcy\u00fc y\u00fckle\nimage_path = 'ornek_insanlar.jpg' # Kendi resminizi buraya koyun\nimg = cv2.imread(image_path)\n\nif img is None:\n    print(f\"Hata: {image_path} bulunamad\u0131 veya y\u00fcklenemedi.\")\nelse:\n    # Y\u00fcz alg\u0131lama yap (burada 'person' s\u0131n\u0131f\u0131n\u0131 hedefliyoruz, y\u00fcz i\u00e7in \u00f6zel model daha iyi)\n    results = model(img)\n\n    # Alg\u0131lanan nesneleri (varsay\u0131lan olarak 'person' s\u0131n\u0131f\u0131) \u00e7iz\n    for r in results:\n        boxes = r.boxes\n        for box in boxes:\n            x1, y1, x2, y2 = map(int, box.xyxy[0])\n            conf = box.conf[0]\n            cls = int(box.cls[0])\n            class_name = model.names[cls]\n\n            # E\u011fer alg\u0131lanan nesne 'person' ise ve g\u00fcven skoru y\u00fcksekse\n            # Y\u00fcz alg\u0131lama i\u00e7in, 'face' s\u0131n\u0131f\u0131 kontrol edilmelidir.\n            if class_name == 'person' and conf > 0.5: \n                cv2.rectangle(img, (x1, y1), (x2, y2), (0, 255, 0), 2)\n                cv2.putText(img, f'{class_name} {conf:.2f}', (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.9, (0, 255, 0), 2)\n\n    # Sonucu g\u00f6ster\n    cv2.imshow('YOLO Yuz Algilama', img)\n    cv2.waitKey(0)\n    cv2.destroyAllWindows()\n<\/pre>\n
<\/code><\/p>\n
Haar Cascades ve YOLO Kar\u015f\u0131la\u015ft\u0131rmas\u0131<\/h4>\n
A\u015fa\u011f\u0131daki tablo, Haar Cascades ve YOLO aras\u0131ndaki temel farklar\u0131 \u00f6zetlemektedir:<\/p>\n\n\n\n\n\n\n\n\n\n\n\n\n\n
\u00d6zellik<\/th>\nHaar Cascades<\/th>\nYOLO (Derin \u00d6\u011frenme Tabanl\u0131)<\/th>\n<\/tr>\n<\/thead>\n
Yakla\u015f\u0131m<\/strong><\/td>\nGeleneksel Makine \u00d6\u011frenimi (\u00d6zellik Tabanl\u0131)<\/td>\nDerin \u00d6\u011frenme (Evri\u015fimli Sinir A\u011flar\u0131)<\/td>\n<\/tr>\n
H\u0131z<\/strong><\/td>\nOrta-Y\u00fcksek (Basit sahnelerde iyi)<\/td>\n\u00c7ok Y\u00fcksek (Ger\u00e7ek zamanl\u0131 performans)<\/td>\n<\/tr>\n
Do\u011fruluk<\/strong><\/td>\nD\u00fc\u015f\u00fck-Orta (Ko\u015fullara duyarl\u0131)<\/td>\n\u00c7ok Y\u00fcksek (Sa\u011flam ve \u00e7e\u015fitli ko\u015fullarda iyi)<\/td>\n<\/tr>\n
Sa\u011flaml\u0131k<\/strong><\/td>\nD\u00fc\u015f\u00fck (Ayd\u0131nlatma, pozisyona duyarl\u0131)<\/td>\nY\u00fcksek (De\u011fi\u015fen ko\u015fullara adaptif)<\/td>\n<\/tr>\n
Yanl\u0131\u015f Pozitifler<\/strong><\/td>\nY\u00fcksek<\/td>\nD\u00fc\u015f\u00fck<\/td>\n<\/tr>\n
E\u011fitim Verisi<\/strong><\/td>\nK\u00fc\u00e7\u00fck, manuel \u00f6zellik m\u00fchendisli\u011fi<\/td>\nB\u00fcy\u00fck veri setleri, otomatik \u00f6zellik \u00f6\u011frenimi<\/td>\n<\/tr>\n
Donan\u0131m \u0130htiyac\u0131<\/strong><\/td>\nD\u00fc\u015f\u00fck (CPU yeterli)<\/td>\nY\u00fcksek (GPU \u00f6nerilir)<\/td>\n<\/tr>\n
Karma\u015f\u0131kl\u0131k<\/strong><\/td>\nD\u00fc\u015f\u00fck (Uygulama)<\/td>\nOrta-Y\u00fcksek (E\u011fitim ve optimizasyon)<\/td>\n<\/tr>\n
Geli\u015fim H\u0131z\u0131<\/strong><\/td>\nDura\u011fan<\/td>\nAktif ve h\u0131zl\u0131 geli\u015fim<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n
Performans Optimizasyonu ve Da\u011f\u0131t\u0131m<\/h4>\n
YOLO modelinizi \u00fcretim ortam\u0131na da\u011f\u0131t\u0131rken performans\u0131 optimize etmek \u00f6nemlidir:<\/p>\n
*   Model Boyutu:<\/strong> YOLOv8'in nano (n<\/code>) veya small (s<\/code>) gibi daha k\u00fc\u00e7\u00fck modellerini kullanarak daha h\u0131zl\u0131 \u00e7\u0131kar\u0131m elde edebilirsiniz.
\n*   Donan\u0131m H\u0131zland\u0131rma:<\/strong> GPU kullan\u0131m\u0131, model \u00e7\u0131kar\u0131m h\u0131z\u0131n\u0131 kat kat art\u0131r\u0131r. NVIDIA GPU'lar i\u00e7in CUDA ve cuDNN, modelin PyTorch veya TensorFlow ile GPU \u00fczerinde \u00e7al\u0131\u015fmas\u0131n\u0131 sa\u011flar.
\n*   Model Optimizasyonu:<\/strong> Modeli ONNX, TensorRT gibi formatlara d\u00f6n\u00fc\u015ft\u00fcrerek donan\u0131ma \u00f6zel optimizasyonlar yapabilirsiniz. Bu, \u00e7\u0131kar\u0131m h\u0131z\u0131n\u0131 daha da art\u0131r\u0131r.
\n*   Batch \u0130\u015fleme:<\/strong> Birden fazla g\u00f6r\u00fcnt\u00fcy\u00fc ayn\u0131 anda i\u015fleyerek (batch processing) GPU kullan\u0131m\u0131n\u0131 optimize edebilirsiniz.
\n*   \u00c7\u0131kar\u0131m K\u00fct\u00fcphaneleri:<\/strong> OpenCV<\/code>'nin DNN<\/code> mod\u00fcl\u00fc veya do\u011frudan ultralytics<\/code> k\u00fct\u00fcphanesi, \u00e7\u0131kar\u0131m i\u00e7in kullan\u0131labilir.<\/p>\n
Ge\u00e7i\u015f Sonras\u0131 De\u011ferlendirme ve Gelecek Trendler<\/h3>\n
Yeni YOLO tabanl\u0131 sisteminizi devreye ald\u0131ktan sonra, performans\u0131n\u0131 s\u00fcrekli olarak izlemek ve gelecekteki geli\u015fmeleri takip etmek \u00f6nemlidir.<\/p>\n
Yeni Sistemin Performans\u0131n\u0131 \u00d6l\u00e7me<\/h4>\n
Ge\u00e7i\u015f sonras\u0131, Haar Cascades ile yapt\u0131\u011f\u0131n\u0131z gibi, yeni sisteminizin performans metriklerini (FPS, do\u011fruluk, yanl\u0131\u015f pozitif\/negatifler) tekrar \u00f6l\u00e7\u00fcn. Bu \u00f6l\u00e7\u00fcmleri farkl\u0131 senaryolarda (d\u00fc\u015f\u00fck \u0131\u015f\u0131k, kalabal\u0131k, farkl\u0131 a\u00e7\u0131lar) yaparak, Haar Cascades'e g\u00f6re ne kadar iyile\u015fme sa\u011flad\u0131\u011f\u0131n\u0131z\u0131 somut verilerle ortaya koyun. A\/B testleri veya eski ve yeni sistemleri paralel \u00e7al\u0131\u015ft\u0131rma gibi y\u00f6ntemlerle kar\u015f\u0131la\u015ft\u0131rmal\u0131 analizler yapabilirsiniz.<\/p>\n
\u00d6l\u00e7eklenebilirlik ve Bak\u0131m<\/h4>\n
YOLO tabanl\u0131 sistemler, derin \u00f6\u011frenme \u00e7er\u00e7evelerinin sa\u011flad\u0131\u011f\u0131 mod\u00fcler yap\u0131 sayesinde daha \u00f6l\u00e7eklenebilirdir. Yeni bir y\u00fcz tipi (\u00f6rne\u011fin maskeli y\u00fczler) i\u00e7in modelinizi yeniden e\u011fitmek veya ince ayar yapmak, Haar Cascades'e g\u00f6re \u00e7ok daha kolayd\u0131r. Ayr\u0131ca, model g\u00fcncellemeleri ve bak\u0131m s\u00fcre\u00e7leri, modern derin \u00f6\u011frenme ara\u00e7lar\u0131 ve platformlar\u0131 sayesinde daha verimli y\u00f6netilebilir. Bulut tabanl\u0131 \u00e7\u00f6z\u00fcmler (AWS SageMaker, Google AI Platform) veya container teknolojileri (Docker, Kubernetes) ile da\u011f\u0131t\u0131m ve \u00f6l\u00e7eklendirme kolayla\u015f\u0131r.<\/p>\n
Y\u00fcz Alg\u0131lamada Gelecek Trendler ve Geli\u015fmeler<\/h4>\n
Y\u00fcz alg\u0131lama alan\u0131 h\u0131zla geli\u015fmeye devam ediyor:<\/p>\n
*   Daha K\u00fc\u00e7\u00fck ve H\u0131zl\u0131 Modeller:<\/strong> G\u00f6m\u00fcl\u00fc sistemler ve mobil cihazlar i\u00e7in daha hafif, enerji verimli modeller geli\u015ftiriliyor.
\n*   Gizlilik Odakl\u0131 Y\u00f6ntemler:<\/strong> Y\u00fcz alg\u0131laman\u0131n gizlilik endi\u015feleri yaratmas\u0131 nedeniyle, gizlili\u011fi koruyan (\u00f6rne\u011fin, y\u00fcz verilerini anonimle\u015ftiren) alg\u0131lama y\u00f6ntemleri \u00fczerine ara\u015ft\u0131rmalar art\u0131yor.
\n*   3D Y\u00fcz Alg\u0131lama:<\/strong> 2D g\u00f6r\u00fcnt\u00fclerdeki s\u0131n\u0131rl\u0131l\u0131klar\u0131 a\u015fmak i\u00e7in 3D sens\u00f6rler ve modeller kullan\u0131larak daha sa\u011flam alg\u0131lama hedefleniyor.
\n*   Transformer Tabanl\u0131 Modeller:<\/strong> Nesne alg\u0131lamada da Transformer mimarileri (DETR gibi) kullan\u0131lmaya ba\u015flanm\u0131\u015f olup, gelecekte y\u00fcz alg\u0131lamada da etkili olmalar\u0131 beklenmektedir.
\n*   Etik ve Adil Yapay Zeka:<\/strong> Y\u00fcz alg\u0131lama sistemlerinin farkl\u0131 demografik gruplar \u00fczerinde adil ve tarafs\u0131z performans g\u00f6stermesi i\u00e7in algoritmik \u00f6nyarg\u0131lar\u0131 azaltmaya y\u00f6nelik \u00e7al\u0131\u015fmalar b\u00fcy\u00fck \u00f6nem ta\u015f\u0131maktad\u0131r.<\/p>\n
Sonu\u00e7<\/h3>\n
Haar Cascades'ten YOLO'ya ge\u00e7i\u015f, y\u00fcz alg\u0131lama sistemlerinizi modernle\u015ftirmenin ve gelece\u011fe haz\u0131rlaman\u0131n \u00f6nemli bir ad\u0131m\u0131d\u0131r. Bu ge\u00e7i\u015f, sadece daha y\u00fcksek do\u011fruluk ve h\u0131z sa\u011flamakla kalmaz, ayn\u0131 zamanda sisteminizin daha sa\u011flam, \u00f6l\u00e7eklenebilir ve \u00e7e\u015fitli uygulama senaryolar\u0131na uyarlanabilir olmas\u0131n\u0131 da sa\u011flar. Derin \u00f6\u011frenmenin sundu\u011fu bu g\u00fc\u00e7l\u00fc ara\u00e7lar\u0131 benimseyerek, y\u00fcz alg\u0131lama projelerinizde yeni ufuklar a\u00e7abilir ve daha geli\u015fmi\u015f \u00e7\u00f6z\u00fcmler sunabilirsiniz. Bu rehberin, bu d\u00f6n\u00fc\u015f\u00fcm yolculu\u011funuzda size yol g\u00f6stermesini umuyoruz.<\/p>\n
SSS (S\u0131k Sorulan Sorular)<\/h3>\n
Neden Haar Cascades'ten YOLO'ya ge\u00e7meliyim?<\/h4>\n
Haar Cascades, eski ve s\u0131n\u0131rl\u0131 bir teknolojidir. YOLO, \u00e7ok daha y\u00fcksek do\u011fruluk, daha iyi ger\u00e7ek zamanl\u0131 performans, de\u011fi\u015fen<\/p>\n","protected":false},"excerpt":{"rendered":"Haar Cascades’ten YOLO’ya: Y\u00fcz Alg\u0131lama Sistemlerini Modernize Etme Rehberi\nBilgisayar g\u00f6r\u00fc\u015f\u00fc alan\u0131nda y\u00fcz alg\u0131lama, y\u0131llard\u0131r temel bir ara\u015ft\u0131…","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"csco_page_header_type":"","csco_page_load_nextpost":"","csco_page_subscribe_form":"","csco_page_contact_form":"","footnotes":""},"categories":[1],"tags":[],"class_list":{"0":"post-39125","1":"post","2":"type-post","3":"status-publish","4":"format-standard","6":"category-genel","7":"cs-entry","8":"cs-video-wrap"},"yoast_head":"\nHaar Cascades'ten YOLO'ya: Y\u00fcz Alg\u0131lama Sistemlerini Modernize Etme Rehberi - Kodlar\u0131n Gizemli D\u00fcnyas\u0131<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/\" \/>\n<meta property=\"og:locale\" content=\"tr_TR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Haar Cascades'ten YOLO'ya: Y\u00fcz Alg\u0131lama Sistemlerini Modernize Etme Rehberi\" \/>\n<meta property=\"og:description\" content=\"Haar Cascades'ten YOLO'ya: Y\u00fcz Alg\u0131lama Sistemlerini Modernize Etme Rehberi Bilgisayar g\u00f6r\u00fc\u015f\u00fc alan\u0131nda y\u00fcz alg\u0131lama, y\u0131llard\u0131r temel bir ara\u015ft\u0131...\" \/>\n<meta property=\"og:url\" content=\"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/\" \/>\n<meta property=\"og:site_name\" content=\"Kodlar\u0131n Gizemli D\u00fcnyas\u0131\" \/>\n<meta property=\"article:published_time\" content=\"2026-02-14T22:00:42+00:00\" \/>\n<meta name=\"author\" content=\"Fatih Soysal\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Yazan:\" \/>\n\t<meta name=\"twitter:data1\" content=\"Fatih Soysal\" \/>\n\t<meta name=\"twitter:label2\" content=\"Tahmini okuma s\u00fcresi\" \/>\n\t<meta name=\"twitter:data2\" content=\"17 dakika\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/#article\",\"isPartOf\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/\"},\"author\":{\"name\":\"Fatih Soysal\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1\"},\"headline\":\"Haar Cascades’ten YOLO’ya: Y\u00fcz Alg\u0131lama Sistemlerini Modernize Etme Rehberi\",\"datePublished\":\"2026-02-14T22:00:42+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/\"},\"wordCount\":2894,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1\"},\"inLanguage\":\"tr\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/#respond\"]}],\"copyrightYear\":\"2026\",\"copyrightHolder\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#organization\"}},{\"@type\":\"WebPage\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/\",\"url\":\"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/\",\"name\":\"Haar Cascades'ten YOLO'ya: Y\u00fcz Alg\u0131lama Sistemlerini Modernize Etme Rehberi - Kodlar\u0131n Gizemli D\u00fcnyas\u0131\",\"isPartOf\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#website\"},\"datePublished\":\"2026-02-14T22:00:42+00:00\",\"breadcrumb\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/#breadcrumb\"},\"inLanguage\":\"tr\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Anasayfa\",\"item\":\"https:\/\/fatihsoysal.com\/blog\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Haar Cascades’ten YOLO’ya: Y\u00fcz Alg\u0131lama Sistemlerini Modernize Etme Rehberi\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#website\",\"url\":\"https:\/\/fatihsoysal.com\/blog\/\",\"name\":\"Fatihsoysal.com\",\"description\":\"Blog - Yaz\u0131l\u0131m D\u00fcnyas\u0131 Tecr\u00fcbelerim\",\"publisher\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/fatihsoysal.com\/blog\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"tr\"},{\"@type\":[\"Person\",\"Organization\"],\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1\",\"name\":\"Fatih Soysal\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"tr\",\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/fatihsoysal.com\/blog\/wp-content\/uploads\/2024\/04\/cropped-replicate-prediction-3kgg1hgjn5rgp0cf0p5tr0jw7w-1.png\",\"contentUrl\":\"https:\/\/fatihsoysal.com\/blog\/wp-content\/uploads\/2024\/04\/cropped-replicate-prediction-3kgg1hgjn5rgp0cf0p5tr0jw7w-1.png\",\"width\":512,\"height\":512,\"caption\":\"Fatih Soysal\"},\"logo\":{\"@id\":\"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/image\/\"},\"description\":\"Kullan\u0131m ve kodlama m\u00fckemmeliyetini odak alan uygulamalar olu\u015fturma deneyimine sahip, profesyonel olarak 15+ y\u0131l \u00fczeri deneyime sahip bir yaz\u0131l\u0131m m\u00fchendisi.\",\"url\":\"https:\/\/fatihsoysal.com\/blog\/author\/fatihsoysal\/\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"Haar Cascades'ten YOLO'ya: Y\u00fcz Alg\u0131lama Sistemlerini Modernize Etme Rehberi - Kodlar\u0131n Gizemli D\u00fcnyas\u0131","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/","og_locale":"tr_TR","og_type":"article","og_title":"Haar Cascades'ten YOLO'ya: Y\u00fcz Alg\u0131lama Sistemlerini Modernize Etme Rehberi","og_description":"Haar Cascades'ten YOLO'ya: Y\u00fcz Alg\u0131lama Sistemlerini Modernize Etme Rehberi Bilgisayar g\u00f6r\u00fc\u015f\u00fc alan\u0131nda y\u00fcz alg\u0131lama, y\u0131llard\u0131r temel bir ara\u015ft\u0131...","og_url":"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/","og_site_name":"Kodlar\u0131n Gizemli D\u00fcnyas\u0131","article_published_time":"2026-02-14T22:00:42+00:00","author":"Fatih Soysal","twitter_card":"summary_large_image","twitter_misc":{"Yazan:":"Fatih Soysal","Tahmini okuma s\u00fcresi":"17 dakika"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/#article","isPartOf":{"@id":"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/"},"author":{"name":"Fatih Soysal","@id":"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1"},"headline":"Haar Cascades’ten YOLO’ya: Y\u00fcz Alg\u0131lama Sistemlerini Modernize Etme Rehberi","datePublished":"2026-02-14T22:00:42+00:00","mainEntityOfPage":{"@id":"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/"},"wordCount":2894,"commentCount":0,"publisher":{"@id":"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1"},"inLanguage":"tr","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/#respond"]}],"copyrightYear":"2026","copyrightHolder":{"@id":"https:\/\/fatihsoysal.com\/blog\/#organization"}},{"@type":"WebPage","@id":"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/","url":"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/","name":"Haar Cascades'ten YOLO'ya: Y\u00fcz Alg\u0131lama Sistemlerini Modernize Etme Rehberi - Kodlar\u0131n Gizemli D\u00fcnyas\u0131","isPartOf":{"@id":"https:\/\/fatihsoysal.com\/blog\/#website"},"datePublished":"2026-02-14T22:00:42+00:00","breadcrumb":{"@id":"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/#breadcrumb"},"inLanguage":"tr","potentialAction":[{"@type":"ReadAction","target":["https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/fatihsoysal.com\/blog\/haar-cascadesten-yoloya-yuz-algilama-sistemlerini-modernize-etme-rehberi\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Anasayfa","item":"https:\/\/fatihsoysal.com\/blog\/"},{"@type":"ListItem","position":2,"name":"Haar Cascades’ten YOLO’ya: Y\u00fcz Alg\u0131lama Sistemlerini Modernize Etme Rehberi"}]},{"@type":"WebSite","@id":"https:\/\/fatihsoysal.com\/blog\/#website","url":"https:\/\/fatihsoysal.com\/blog\/","name":"Fatihsoysal.com","description":"Blog - Yaz\u0131l\u0131m D\u00fcnyas\u0131 Tecr\u00fcbelerim","publisher":{"@id":"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/fatihsoysal.com\/blog\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"tr"},{"@type":["Person","Organization"],"@id":"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/002a254750921dcfd568a99e48240dd1","name":"Fatih Soysal","image":{"@type":"ImageObject","inLanguage":"tr","@id":"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/image\/","url":"https:\/\/fatihsoysal.com\/blog\/wp-content\/uploads\/2024\/04\/cropped-replicate-prediction-3kgg1hgjn5rgp0cf0p5tr0jw7w-1.png","contentUrl":"https:\/\/fatihsoysal.com\/blog\/wp-content\/uploads\/2024\/04\/cropped-replicate-prediction-3kgg1hgjn5rgp0cf0p5tr0jw7w-1.png","width":512,"height":512,"caption":"Fatih Soysal"},"logo":{"@id":"https:\/\/fatihsoysal.com\/blog\/#\/schema\/person\/image\/"},"description":"Kullan\u0131m ve kodlama m\u00fckemmeliyetini odak alan uygulamalar olu\u015fturma deneyimine sahip, profesyonel olarak 15+ y\u0131l \u00fczeri deneyime sahip bir yaz\u0131l\u0131m m\u00fchendisi.","url":"https:\/\/fatihsoysal.com\/blog\/author\/fatihsoysal\/"}]}},"yoast_meta":{"yoast_wpseo_title":"","yoast_wpseo_metadesc":"","yoast_wpseo_canonical":""},"amp_enabled":true,"_links":{"self":[{"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/posts\/39125","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/comments?post=39125"}],"version-history":[{"count":0,"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/posts\/39125\/revisions"}],"wp:attachment":[{"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/media?parent=39125"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/categories?post=39125"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/fatihsoysal.com\/blog\/wp-json\/wp\/v2\/tags?post=39125"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}