An Implementation of Face Recognition with Deep Learning based on a Container-Orchestration Platform

As a lightweight alternative to a virtual machine, a container runs applications only with the necessary environmental variables, libraries, etc. Moreover, many more containers can be run on the same computer compared to traditional VMs, which take up a lot of computing resources. Currently, Docker container and Kubernetes (K8s), which is a container-orchestration platform, are very popular tools. In addition, K8s is a high availability (HA) system with many features that can provide containers to implement more applications. In this project, a face recognition application is implemented with deep learning on Kubeflow, which is a machine learning platform running on K8s. Also, the deep learning method output features instead of classifications. This method computes the distance between two images with Triplet loss function and Euclidean distance. K8s runs on the server as a private cloud, on which our face recognition application runs.