search

anilsathyan7 / Portrait-Segmentation

Real-time portrait segmentation for mobile devices
MIT License
638 stars 133 forks source link

An Issue when loading SlimNet model on Mobile GPU #44

Open rose-jinyang opened 3 years ago

rose-jinyang commented 3 years ago

Hello How are you? Thanks for contributing to this project. I trained a model with SlimNet architecture on my dataset and got a TFlite model. This TFlite model works well on the CPU. I am going to use this TFLite model on mobile GPU. When loading this model on Android GPU, I got the following issue.

Caused by: java.lang.IllegalArgumentException: Internal error: Failed to apply delegate: Attempting to use a delegate that only supports static-sized tensors with a graph that has dynamic-sized tensors.

So I searched for any solution but I did NOT find a correct solution yet. https://github.com/tensorflow/tensorflow/issues/38036

Could u help me?

anilsathyan7 commented 3 years ago

Can you try the following code to fix the size of the model during conversion

# Fix the model input batch size to 1
input_name = float_model.input_names[0]
index = float_model.input_names.index(input_name)
float_model.inputs[index].set_shape([1, 512, 512, 3])

converter = tf.lite.TFLiteConverter.from_keras_model(float_model)
tflite_model = converter.convert()
open("slimnet_float.tflite", "wb").write(tflite_model)
rose-jinyang commented 3 years ago

But I met the following issue.

image

Can u run the slimnet model on Phone GPU?

anilsathyan7 commented 3 years ago

I tested it with mediapipe api in android(segmentation) and it worked at that time. I believe it uses the old gpu-delegate with OpenGL instead of OpenCL.