cengizhanyurdakul / kubeface-protoplayground

Exploring Kubernetes, Face Detection, and gRPC

Kubernetes + Face Detection + gRPC

Installation

Minikube

We usually use Minikube to control cluster and other components. The minikube can be installed quickly and practically from the link.

Kubernetes

You can follow the installations in the link to do Kubernetes operations.

gRPC

It is currently installed inside the requirements.txt file. You can easily install gRPC with package install.

pip install grpcio grpcio-tools

If you make changes to the inputs and outputs in the proto files, do not forget to recreate the service files with the following command.

python -m grpc_tools.protoc -I src/protos/ --python_out=src/gRPCs/ --grpc_python_out=src/gRPCs/ src/protos/detection.proto

RUN

With the run.sh file, all processes can become available. In the file, the purpose of each command is specified. In summary, docker image is built, Kubernetes cluster is run, the built image is loaded to the cluster, kubernetes config is applied and finally service information is given.

bash run.sh

Once the build is up and running, you can detect all the photos in the InputImages file with the following command.

python Client.py

With the minikube dashboard command, you can monitor the status of services and other components on the page that will open.

minikube dashboard

Modifiable Parameters

• replicas: Parameter that build a specific number of pods (Face Detector Service)
• cpus and memory: Information on how much cpu and memory the cluster will use when running

References

• https://github.com/hukkelas/DSFD-Pytorch-Inference