Two different datasets have been considered during the experiments:
Labeled Faces in the Wild
The Labeled Faces in the Wild (LFW) [14] is composed by 13, 233 face images (10, 256 males and 2977 females) from 5749 celebrities collected from the web. The Facial Image Processing and Analysis group (FIPA, http://fipa.cs.kit.edu) proposed a benchmark to evaluate the gender recognition problem using these images. Following this protocol, we have employed the 5 folds described in [9], using 4 folds for training and 1 for testing, and the results were averaged. Similar articles in the bibliography have used a reduced version of this dataset, taking out a huge number of challenging images [27,37]. However, it is important to note that our experiments are performed using the entire dataset, without discarding any image.
Two different datasets have been considered during the experiments:Labeled Faces in the Wild The Labeled Faces in the Wild (LFW) [14] is composed by 13, 233 face images (10, 256 males and 2977 females) from 5749 celebrities collected from the web. The Facial Image Processing and Analysis group (FIPA, http://fipa.cs.kit.edu) proposed a benchmark to evaluate the gender recognition problem using these images. Following this protocol, we have employed the 5 folds described in [9], using 4 folds for training and 1 for testing, and the results were averaged. Similar articles in the bibliography have used a reduced version of this dataset, taking out a huge number of challenging images [27,37]. However, it is important to note that our experiments are performed using the entire dataset, without discarding any image.
การแปล กรุณารอสักครู่..