年齡估測的相關研究是電腦視覺領域中的熱門主題之一，因為年齡識別可以廣泛地應用於生活中。例如：(1)消費場所的管理：具有年齡辨識功能的設備，可以用於自動過濾來售出有年齡限制的產品（如菸、酒、電影）或是在有年齡限制的場所的管制人員出入；（2）商業策略的設定：在商業銷售上，不同年齡層的購物習慣與喜好大不相同，自動年齡數據的蒐集可以提供業者相關資訊進行市場分析，如電子化顧客關係管理 (ECRM); (3)辨識系統效能的提升：年齡也是一種生物特徵，可用於協助主要生物特徵來提高人物的recognition，verification 或是authentication 等應用的正確率。
隨著深度學習被廣泛應用於視覺領域，年齡估測的準確度也不斷提升。目前很多研究利用深度學習提出估測年齡的方法，較早的研究其系統深度僅四到五層，後來系統深度漸漸加大，也提高了準確度。因此本論文提出一個基於CNN深度學習架構的年齡估測系統。此系統改良自DEX系統，採用VGG16為系統核心，並使用多重損失函數進行訓練，此多重損失函數考量softmax, 平均數與變異數三種損失。實驗除了採用ChaLearn LAP(2015)的資料庫及其評比方式，也在AFAD與MORPH II兩組年齡資料庫下進行測試。實驗結果說明本系統在使用IMD-WIKI的預訓練下，外表年齡估測方面的表現有較其他深度學習架構的結果來的好，並且在實際年齡的資料庫(MORPH II、AFAD)上，也有較好的結果。

Age estimation has been one of hot topics in computer vision. Identifying personal characteristics such as age, personal identity, gender, and ethnicity through images is an interesting but challenging problem. In recent years, age estimation has become an attractive research topic because it can be widely applied to human life. For example: (1) devices with age recognition can automatically filter age-restricted products, such as cigarettes and alcohol. (2) Since shopping habit and preferences of different age groups are very different, the automatic collection of age data can provide relevant information for market analysis, such as Electronic customer relationship management (ECRM). (3) Age is a biological feature that can be used to assist major biometrics to improve the accuracy of human recognition, verification or authentication applications.
As deep learning is widely used in the computer vision, the accuracy of age estimation is also increasing. Early CNN-based works used four to five layers for depth; however, current works adopts a deeper structure and it results in more accurate results. This thesis proposes an age estimation system based on CNN deep learning architecture. The proposed system is modified from DEX system of which VGG16 is adopted as the learning core. We incorporate a multi-loss function which takes into accout the losses of softmax, mean and variance. In the experiments, we show the performance on the apparent database ChaLearn LAP (2015). Moreover, the proposed system is also tested on the real age databases, i.e. AFAD and MORPH II. Experimental results show that the system has a better performance in both the apparent and real age databases.

Chapter 1 Introduction 1
1.1 Motivation 1
1.2 System Overview 2
1.3 Thesis Organization 3
Chapter 2 Related Work 4
2.1 Biological age estimation 4
2.2 Apparent age estimation 5
Chapter 3 System Description 7
3.1 Face extraction from wild images 7
3.2 Age Estimation System with a Multi-Loss Function 8
3.2.1 Description of VGG16 8
3.2.2 Multi-Loss Function 10
Chapter 4 Experimental Results 12
4.1 Aging training databases 12
4.2 Evaluation metric 15
4.3 Experimental results and analysis 16
4.3.1 Weights of the multi-loss function 16
4.3.2 AFAD database age estimation 19
4.3.3 Real age estimation 20
Chapter 5 Conclusion 22
Reference 23

[1]X. Geng, Z.H. Zhou, and K. Smith-Miles, “Automatic Age Estimation Based on Facial Aging Patterns,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 29, no. 12, pp. 2234-2240, Dec. 2007.
[2]K.Y. Chang, C.S. Chen, and Y.P. Hung, “Ordinal Hyperplanes Ranker with Cost Sensitivities for Age Estimation,” in Proc. IEEE Conf. Computer Vision and Pattern Recognition, pp. 585-592, 2011.
[3]K. Chen, S.-G. Gong, T. Xiang, and C.-C. Loy, “Cumulative Attribute Space for Age and Crowd Density Estimation,” in Proc.IEEE Conf. Computer Vision and Pattern Recognition, pp. 2467-2474, 2013.
[4]G. Guo, G. Mu, and A. Graves, “Joint Estimation of Age, Gender and Ethnicity,” in Proc. International Conference on Automatic Face and Gesture Recognition, pp.1-6, 2013.
[5]X. Wang, R. Guo, and C. Kambhamettu, “Deeply-Learned Feature for Age Estimation,” in Proc. IEEE Winter Conf. on Applications of Computer Vision, pp.534-541, 2015.
[6]G. Levi, and T. Hassner, “Age and Gender Classification Using Convolutional Neural Networks,” in Proc. IEEE Conference on Computer Vision and Pattern Recognition, pp. 34–42, 2015.
[7]D. Yi, Z. Lei, and S Z. Li, “Age Estimation by Multi-Scale Convolutional Network,” in Proc. Asian Conference on Computer Vision, pp. 144–158, 2014.
[8]L. Hou, D. Samaras, and T M. Kurc, “Neural Networks with Smooth Adaptive Activation Functions for Regression,” arXiv preprint, arXiv:1608.06557, 2016.
[9]S. Escalera, M. Torres Torres, B. Mart´ınez, X. Baro, H. J. Escalante, I. Guyon, G. Tzimiropoulos, C. Corneanu, M. Oliu, M. A. Bagheri, and M. Valstar, “ChaLearn Looking at People and Faces of the World: Face Analysis Workshop and Challenge 2016,” in Proc. IEEE Conf. on Computer Vision and Pattern Recognition Workshops, pp.706-713, 2016.
[10]X. Liu, S. Li, M. Kan, J. Zhang, S. Wu, W. Liu, H. Han, S. Shan, and X. Chen, “AgeNet: Deeply Learned Regressor and Classifier for Robust Apparent Age Estimation,” in Proc. IEEE Inter. Conf. on Computer Vision Workshops, 2015.
[11]G. Antipov, M. Baccouche, S.A. Berrani, and J.L. Dugelay, “Apparent Age Estimation from Face Images Combining General and Children-Specialized Deep Learning Models,” in Proc.IEEE Conf. on Computer Vision and Pattern Recognition Workshops, pp.706-713, 2016.
[12]M. Mathias, R. Benenson, M. Pedersoli, and L. V. Gool, “Face Detection without Bells and Whistles,” in Proc. European Conference on Computer Vision, pp. 720-735, 2014.
[13]Z. Huo, X. Yang, C. Xing, Y. Zhou, P. Hou, J. Lv, and X. Geng, “Deep Age Distribution Learning for Apparent Age Estimation,” in Proc. IEEE Conf. on Computer Vision and Pattern Recognition Workshops, pp.722-729, 2016.
[14]K. Simonyan, and A. Zisserman, “Very Deep Convolutional Networks for Large-Scale Image Recognition,” in Proc. International Conference on Learning Representations 2015, 2015.
[15]J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and F.-F. Li, “ImageNet: A Large-Scale Hierarchical Image Database,” in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, pp. 248–255, Jun. 2009.
[16]A. Krizhevsky, I. Sutskever, and G. E. Hinton “ImageNet Classification with Deep Convolutional Neural Networks,” in Proc. Conference on Neural Information Processing Systems, pp. 1097-1105, 2012.
[17]H. Pan, H. Han, S. Shan, and X. Chen, “Mean-Variance Loss for Deep Age Estimation from a Face,” in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, pp.5285-5294, 2018.
[18]R. Rothe, R. Timofte, and L.V. Gool, “DEX: Deep Expectation of Apparent Age from a Single Image,” in Proc. Inter. Conf. on Computer Vision, pp.252-257, 2015.
[19]R. Rothe, R. Timofte, and L.V. Gool, “Deep Expectation of Real and Apparent Age from a Single Image Without Facial Landmarks,” International Journal of Computer Vision, vol.126, no.2-4, pp.144-157, 2018.
[20]S. Escalera, J. Fabian, P. Pardo, X. Baro, J. Gonzalez, H. J. Escalante, D. Misevic, U. Steiner, and I. Guyon, “ChaLearn Looking at People 2015: Apparent Age and Cultural Event Recognition Datasets and Results,” in Proc. IEEE Conf. on International Conference on Computer Vision Workshop, pp. 1714–1722, 2015.
[21]O. Russakovsky, J. Deng, H. Su, J. Krause, S. Satheesh, S. Ma, Z. Huang, A. Karpathy, A. Khosla, M. Bernstein, A. C. Berg, and F –F Li, “ImageNet Large Scale Visual Recognition Challenge,” International Journal of Computer Vision, vol.115, no.3, 2014.
[22]K. Ricanek, and T. Tesafaye, “MORPH: a Longitudinal Image Database of Normal Adult Age-Progression,” in Proc. International Conference on Automatic Face and Gesture Recognition, 2006.
[23]Z. Niu, M. Zhou, L. Wang, X. Gao, and G. Hua, “Ordinal Regression with Multiple Output CNN for Age Estimation,” in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, pp. 4920–4928, 2016.
[24]K. Chang, C. Chen, S. Shan, and Y. Hung, “Ordinal Hyperplanes Ranker with Cost Sensitivities for Age Estimation,” in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, pp. 585– 592, 2011.