Kubeflow pipeline is inherently laborious to construct. It involves procedures such as docker file creation and compilation into docker image and yaml file for deployment on kubwflow suite for training and deployment. For an AIOT flow to function rapidly for results evaluation and prototyping, a drag to drop GUI front end is devised in our proposal for model deployment with Rasberry PI IOT sensing capability. The GUI is implemented through a Node-red open source development suite which features drag and drop and custom node capabilities.
The application scenario first starts from a trained model deployment through Kserve model server. Then a pipeline is constructed that take an input from Raberry PI sensor, then feed into the model server through kserve API. An output of the model testing is generated by the model server and fed into the PI chipset for responses. All these integrations including kserve model pipeline and Rasberry PI I/O will be seen on the Node-red custom pallet and canvas. Please let us know, if they are of interest to this repository. The code is ready for PR.
Kubeflow pipeline is inherently laborious to construct. It involves procedures such as docker file creation and compilation into docker image and yaml file for deployment on kubwflow suite for training and deployment. For an AIOT flow to function rapidly for results evaluation and prototyping, a drag to drop GUI front end is devised in our proposal for model deployment with Rasberry PI IOT sensing capability. The GUI is implemented through a Node-red open source development suite which features drag and drop and custom node capabilities. The application scenario first starts from a trained model deployment through Kserve model server. Then a pipeline is constructed that take an input from Raberry PI sensor, then feed into the model server through kserve API. An output of the model testing is generated by the model server and fed into the PI chipset for responses. All these integrations including kserve model pipeline and Rasberry PI I/O will be seen on the Node-red custom pallet and canvas. Please let us know, if they are of interest to this repository. The code is ready for PR.