Predict via TensorFlow Serving (gRPC)

This component takes an image column as input and reachs out to a configured TensorFlow Serving instance via gRPC to get predictions for the image. Output of the component is the resulting image. If you do not expect an image as output but, e.g., one or more double values, use the "Image to Data Row" node afterwards for conversion.

Note: When using this component to create a workflow that should run in KNIME Edge, make sure to have the same model (and version) deployed in KNIME Edge as using locally for testing.

Options

Image input column
The column containing the images to predict.
Model version (-1 means latest)
The version of the specified model to use. If set to -1, the latest available model will be used.
gRPC port (will be ignored by KNIME Edge)
The port of the TensorFlow Serving instance that is used when running this component locally or on KNIME Server. When running this component in KNIME Edge, the port will automatically be replaced, so that the TensorFlow Serving instance running in the Edge cluster will be used.
Model name
The name of the model that is deployed in TensorFlow Serving and should be used for predictions.
Host (will be ignored by KNIME Edge)
The host of the TensorFlow Serving instance that is used when running this component locally or on KNIME Server. When running this component in KNIME Edge, the host will automatically be replaced, so that the TensorFlow Serving instance running in the Edge cluster will be used.
Input layer name
The name of the input layer of the TensorFlow network.
Output layer name (first output taken if empty)
The name of the output layer of the TensorFlow network. This field can be left empty. In that case, the first output (if there are multiple) that is listed by TensorFlow Serving will be used.

Input Ports

Icon
A table with an image column.

Output Ports

Icon
The input table with an appended column that contains the resulting images. Use the "Image to Data Row" node if you expect numbers instead of an image as output.

Nodes

Extensions

Links

Download

To use this component in KNIME, download it from the below URL and open it in KNIME:

Download Component
On NodePit since: 2021-06-09
Last update: 2024-04-27
Created with KNIME version: v4.3.2

Deploy, schedule, execute, and monitor your KNIME workflows locally, in the cloud or on-premises – with our brand new NodePit Runner.

Try NodePit Runner!