YiuzBulma Yontemlerilnln Haber V'ideolari IClnSIstemat'ik KarE1atird1masi Systematic Evaluation ofFaceDetection Algorithms on NewsVideos

People arethemostimportant subjects innewsvideos andfor proper retrieval ofpeople images; face detection isaverycru- cial step. However, facedetection andrecognition innewsvi- deosisaverychallenging task duetothehugeirregularities and highnoise level inthedata. Inaddition tothat, withdifferent face detection algorithms, thenumber andthetypeofthefaces maydiffer. Inthis study, inorder togetthebest performance fromexisting methods, systematic evaluation ofthese methods isperformed. Intheexperiments, newsvideos fromTRECVID 2006data setareusedandforevaluation four different face de- tection methods arechosen. 1.Giri,