Failed TensorflowLite Operation

[Zumbalamambo]

The input shape of my model is [ 1 256 256 3]

I'm getting the following errror,

Exception: TensorFlowLite operation failed.
TensorFlowLite.Interpreter.ThrowIfError (System.Int32 resultCode)

How do I fix this?

I have invoked the Model as follows,

model.Invoke(background.texture);

background is the rawimage. I tried to save the raw image to png image and ocourse, it is not null.

Following is my Invoke function,

public override void Invoke(Texture inputTex)
        {
            ToTensor(inputTex, inputs);

            Debug.LogWarning(inputTex.height + " " + inputs.Length);
            interpreter.SetInputTensorData(0,  inputs);
            interpreter.Invoke();
        }

How do I fix this error?

[asus4]

I guess its the tensor format mismatching. you can check the input format with the following method. or Netron is a good GUI app.

var info = interpreter.GetInputTensorInfo(index)

or could you share the tflite file? It is difficult to identify this type of program w/o the model.

[Zumbalamambo]

i tried Netron. This is the input part of the graph.

[asus4]

Click input, then detailed information appears.
MNIST's input is float32, SSD is quantized to uint8.

[kasi-str8bat]

Hi, can i turnoff the quantization from the library? is it possible to import float32 model?

[tklecka]

@asus4
Same problem here.
The model works well in the Android demo.
The shape is [1,300,300,3].
model.zip

      at TensorFlowLite.Interpreter.ThrowIfError (TensorFlowLite.Interpreter+Status status) [0x00000] in <00000000000000000000000000000000>:0 
      at TensorFlowLite.Interpreter.SetInputTensorData (System.Int32 inputTensorIndex, System.Array inputTensorData) [0x00000] in <00000000000000000000000000000000>:0 
      at TensorFlowLite.SSD.Invoke (UnityEngine.Texture inputTex) [0x00000] in <00000000000000000000000000000000>:0 
       ...

[JoochanKim-LTU]

Hi, can i turnoff the quantization from the library? is it possible to import float32 model?

@asus4
Same problem here.
The model works well in the Android demo.
The shape is [1,300,300,3].
model.zip

      at TensorFlowLite.Interpreter.ThrowIfError (TensorFlowLite.Interpreter+Status status) [0x00000] in <00000000000000000000000000000000>:0 
      at TensorFlowLite.Interpreter.SetInputTensorData (System.Int32 inputTensorIndex, System.Array inputTensorData) [0x00000] in <00000000000000000000000000000000>:0 
      at TensorFlowLite.SSD.Invoke (UnityEngine.Texture inputTex) [0x00000] in <00000000000000000000000000000000>:0 
       ...

I'm also trying to solve this issue. Any luck yet?

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[klauspa]

I can't quantized my SSD model, how do I modify the code so that my 32bit model can be loaded in this project? many thanks