Combined Deep Learning With Directed Acyclic Graph SVM for Local Adjustment of Age Estimation

In order to further improve the accuracy of age estimation, a locally adjusted estimation algorithm based on deep learning and directed acyclic graph SVM is proposed. training phase, SE-ResNet-50 network pre-trained by VGGFace2 dataset first fine-tuned. Once converges, vector consisting parameters last fully connected layer used as representation train multiple One-Versus-One SVMs. test we sent face image be estimated into obtain rough value, then set specific neighborhood, finally combined trained neighborhood with global estimate center for accurate estimate. show universality proposed coarse-to-fine or/and global-to-local method, experiments were carried out MORPH AFAD images different races, results verified effectiveness algorithm.