Sensitivity Evaluation of SNoW and SVM Based Face Detection Methods
Jialin Liu, Mika Rautiainen, Timo Ojala · 2006
This paper studies the performance sensitivity of the variations in different factors on SNoW and SVM based face detection techniques. These factors include combinations of training set, pixel transitions in sliding window, pre-processing methods (rotation, scaling, light correction, masking filter), and bootstrapping in negative training patterns. It was found that performance of the two detection methods is significantly sensitive to these factors.