Help Needed, NullReferenceException when calling runner.AddInput()

[andyl1992]

I am trying to infer with a pre-trained model i have created with keras/tensorflow. I am adapting code from the examples to try and achieve. I believe i have adapted the ImageUtils class correctly for my model but i am not sure about the rest.

I am getting a NullReferenceException when calling runner.AddInput()

TensorFlow.TFGraph.this[string].get returned null

        var graph = new TFGraph();
        var model = File.ReadAllBytes("220318model321flat.pb");
        graph.Import(model);

        if (graph == null)
        {
            Console.WriteLine("graph empty");
        }

        using (var session = new TFSession(graph))
        {

            var tensor = ImageUtil.CreateTensorFromImageFile("2.png", TFDataType.UInt8);
            Console.WriteLine(tensor.GetTensorDimension(1).ToString() + tensor.GetTensorDimension(2).ToString());

            if (tensor == null)
            {
                Console.WriteLine("tensor empty");
            }

            var runner = session.GetRunner();
            runner.AddInput(graph["input"][0], tensor).Fetch(graph["output"][0]);

            var output = runner.Run();
            var result = output[0];

        }

Any help would be much appreciated, thank you.

[raulbojalil]

This means that the specified operation ("input") does not exist in your graph. You can get a list of all operations by calling graph.GetEnumerator().

[coelhojs]

For future reference, a tutorial from Tensorflow documentation defined the variables input as "Placeholder" and output as "final_result"