tensorflow: TensorFlow demo app crashes with my own model

What related GitHub issues or StackOverflow threads have you found by searching the web for your problem?

https://github.com/tensorflow/tensorflow/issues/1269

Environment info

Operating System: Linux ubuntu

If installed from source, provide

1. The commit hash (git rev-parse HEAD) ce3572a08b9ecfa5c8dd94921c2011f37b58e608

If possible, provide a minimal reproducible example (We usually don’t have time to read hundreds of lines of your code)

1. Followin this tutorial: https://www.tensorflow.org/versions/r0.9/how_tos/image_retraining/index.html I built a model via transfer learning using custom images.
2. I edited WORKSPACE and built demo.apk App crashed.
3. Then i used inception5h.zip as referenced in https://github.com/tensorflow/tensorflow/tree/master/tensorflow/examples/android. The app run without any problem.

What other attempted solutions have you tried?

Then i looked at issue https://github.com/tensorflow/tensorflow/issues/1269 for guidance.

1. —Edited tensorflowImagelistener.java—

  private static final int NUM_CLASSES = 4; //1001;
  private static final int INPUT_SIZE = 299; //224
  private static final int IMAGE_MEAN = 128; //117
  private static final float IMAGE_STD = 128; //1;
  private static final String INPUT_NAME = "Mul:0"; //"input:0"; 
  private static final String OUTPUT_NAME = "final_result:0"; //"output:0";

1. —I even stripped the graph using strip_unused.py using ----

  bazel build tensorflow/python/tools:strip_unused
  bazel-bin/tensorflow/python/tools/strip_unused --input_graph=/tmp/inception.pb   --output_graph=/tmp/stripped_inception.pb --input_node_names="Mul:0" --output_node_names="final_result" --input_binary=true

1. –In strip_unused_lib.py i added if else clause as referenced by @dmirk quick-n-dirty from #1269—

      if "jpeg" in node.op.lower():
        placeholder_node.attr["dtype"].CopyFrom(tf.AttrValue(
            type=tf.uint8.as_datatype_enum))
      else:
        placeholder_node.attr["dtype"].CopyFrom(tf.AttrValue(
            type=placeholder_type_enum))

1. For some reason i still got this error. See attached oldlogcat.txt

tensorflow_jni.cc:304 Error during inference: Invalid argument: No OpKernel was registered to support Op 'DecodeJpeg' with these attrs.  Registered kernels:

1. As a last resort i did this modification to tensorflow_jni.cc as referenced in #1269. See attached logcat.txt

      // Copy 3 values
      input_tensor_mapped(0, i, j, 0) =
          (static_cast<float>(src->red) - g_image_mean) / g_image_mean;//g_image_std;
      input_tensor_mapped(0, i, j, 1) =
          (static_cast<float>(src->green) - g_image_mean) / g_image_mean;//g_image_std;
      input_tensor_mapped(0, i, j, 2) =
          (static_cast<float>(src->blue) - g_image_mean) / g_image_mean;//g_image_std;

but i think this dint help any much. however i dint find the opkernel error

Logs or other output that would be helpful