Age estimation using face images is an exciting and challenging task. The traits from the face images are used to determine age, gender, ethnic background, and emotion of people. Among this set of traits, age estimation can be valuable in several potential real-time applications. The traditional hand-crafted methods relied-on for age estimation, cannot correctly estimate the age. The availability of huge datasets for training and an increase in computational power has made deep learning with convolutional neural network a better method for age estimation; convolutional neural network will learn discriminative feature descriptors directly from image pixels. Several convolutional neural net work approaches have been proposed by many of the researchers, and these have made a significant impact on the results and performances of age estimation systems. In this paper, we present a thorough study of the state-of-the-art deep learning techniques which estimate age from human faces. We discuss the popular convolutional neural network architectures used for age estimation, presents a critical analysis of the performance of some deep learning models on popular facial aging datasets, and study the standard evaluation metrics used for performance evaluations. Finally, we try to analyze the main aspects that can increase the performance of the age estimation system in future.

中文翻译：

---

面部年龄分类的深度学习方法：最先进的调查

使用人脸图像进行年龄估计是一项令人兴奋且具有挑战性的任务。面部图像的特征用于确定人们的年龄、性别、种族背景和情感。在这组特征中，年龄估计在几个潜在的实时应用中可能很有价值。传统的手工方法依赖于年龄估计，无法正确估计年龄。用于训练的庞大数据集的可用性和计算能力的增加使卷积神经网络的深度学习成为更好的年龄估计方法；卷积神经网络将直接从图像像素中学习判别特征描述符。许多研究人员已经提出了几种卷积神经网络工作方法，这些方法对年龄估计系统的结果和性能产生了重大影响。在本文中，我们对最先进的深度学习技术进行了深入研究，该技术可根据人脸估计年龄。我们讨论了用于年龄估计的流行卷积神经网络架构，对一些深度学习模型在流行的面部老化数据集上的性能进行了批判性分析，并研究了用于性能评估的标准评估指标。最后，我们尝试分析未来可以提高年龄估计系统性能的主要方面。并研究用于绩效评估的标准评估指标。最后，我们尝试分析未来可以提高年龄估计系统性能的主要方面。并研究用于绩效评估的标准评估指标。最后，我们尝试分析未来可以提高年龄估计系统性能的主要方面。