**4. Conclusions**

This study was motivated by the fact that training a CNN model based on age comparison is easier than directly estimating the absolute age. The proposed approach trains a CNN model for age comparison using a Siamese network-based deep metric learning method. We designed a binary classifier, which is applied to train a Siamese network to cluster the classes within the margin of tolerance as the same class, allowing us to successfully train the Siamese network by adopting *L*1 − *norm* instead of *L*<sup>2</sup> − *norm*. The experimental test indicated that the proposed approach itself performs the gender classification in processing the age estimation, thus we tried training the CNN model by comparing age and gender simultaneously using the multi-task learning technique. The proposed method was evaluated using the MORPH dataset. Although our architecture needs many epochs to converge, it results in better performance. There was also an additional enhancement using multi-task learning for age and gender compared to CRCNN, the original Siamese network-based deep metric learning, and the latest M-LSDML. Additionally, our method also has better results than the state of the art on MegaAge-Asian and MORPH datasets. In future work, more analysis is needed to reduce the training time by selecting reference images to compare rather than comparing all images.

**Author Contributions:** Y.J. designed the entire core architecture and performed the hardware/software implementation and experiments; S.L. validated the experimental results by the proposed framework; and K.H.P. proposed the key concept and algorithm of the proposed architecture. D.P. is the corresponding author.

**Funding:** This study was supported by the BK21 Plus project funded by the Ministry of Education, Korea (21A20131600011).

**Conflicts of Interest:** The authors declare no conflict of interest. The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.
