Example Tensorflow Processor using Java API for Apache NiFi 1.2+
Example using out of box TensorFlow Java example with NiFi
Article detailing creation, building and usage https://community.hortonworks.com/content/kbentry/116803/building-a-custom-processor-in-apache-nifi-12-for.html
Currently simple classification models are supported. The model directory should have a file graph.pb containing a tensorflow graph, and a label.txt file containing labels by index of the output classes of the graph. The model must also have a variable called input which expects as Image tensor, and a output variable which stores a tensor of label index and probability.
This is a clean update for TensorFlow 1.6. This takes a flow file as an image
JPG, PNG, GIF
Updated TensorFlowProcessor to TF 1.6. Added more tests. More cleanup. Top 5 returned in clean naming.
Install to /usr/hdf/current/nifi/lib/