Age Estimation from Face Images Based on Deep Learning

Age estimation via face images is widely used in different fields, including intelligent security, financial transactions, and public transportation. Even after many years' research, evaluating the specific age from facial images is still a challenging problem. Recently, deep learning-based methods are playing an important role in image processing, which provides a possible way to improve estimation models. We estimate accurate age and age range by testing several deep learning models, which include InceptionResNetV2, Xception, DenseNet121, ResNet152V2, VGG16, and InceptionV3. By comparing their performance, we discover that Xception has the highest accuracy of age estimation. To optimize the performance of these convolutional neural network, we apply the soft labelling technique and analyze the experiment results.