How Can I convert this model to TFLITE

[kei9327]

I want to build Android App.

So I use Tensorflow Lite for Android.

But I can't get .tflite file.

How Can I convert this model to TFLITE?

[kei9327]

@akirasosa please help

[sercant]

Hello @kei9327,

Here is a script that I wrote to convert the model to tf-lite. Also, the model is being tested with the python tf-lite interpreter in the code but you can just comment that part out if you don't need it.

https://gist.github.com/sercant/478cac13391e1b69b2be07654cf3d21e

tf-convert-tflite.py

import argparse

import cv2
import matplotlib.pyplot as plt
import numpy as np
import tensorflow as tf

from data import standardize

prefix = 'hair_recognition'

def main(pb_file, img_file):
    """
    Predict and visualize by TensorFlow.
    :param pb_file:
    :param img_file:
    :return:
    """
    with tf.gfile.GFile(pb_file, "rb") as f:
        graph_def = tf.GraphDef()
        graph_def.ParseFromString(f.read())

    with tf.Graph().as_default() as graph:
        tf.import_graph_def(graph_def, name=prefix)

    for op in graph.get_operations():
        print(op.name)

    x = graph.get_tensor_by_name('%s/input_1:0' % prefix)
    y = graph.get_tensor_by_name('%s/output_0:0' % prefix)

    loaded_image = cv2.cvtColor(cv2.imread(img_file,-1), cv2.COLOR_BGR2RGB)
    resized_image =cv2.resize(loaded_image, (128, 128))
    input_image = np.expand_dims(np.float32(resized_image[:128, :128]),axis=0)/255.0

    # images = np.load(img_file).astype(float)
    # img_h = images.shape[1]
    # img_w = images.shape[2]

    with tf.Session(graph=graph) as sess:
        # for img in images:
        # batched = img.reshape(-1, img_h, img_w, 3)
        normalized = standardize(input_image)

        converter = tf.contrib.lite.TocoConverter.from_session(sess, [x], [y])
        tflite_model = converter.convert()
        open("converted_model.tflite", "wb").write(tflite_model)

        # Load TFLite model and allocate tensors.
        interpreter = tf.contrib.lite.Interpreter(model_content=tflite_model)
        interpreter.allocate_tensors()

        # Get input and output tensors.
        input_details = interpreter.get_input_details()
        output_details = interpreter.get_output_details()

        # Test model on random input data.
        # input_shape = input_details[0]['shape']
        input_data = np.array(normalized, dtype=np.float32)
        interpreter.set_tensor(input_details[0]['index'], input_data)

        interpreter.invoke()
        output_data = interpreter.get_tensor(output_details[0]['index'])
        # print(output_data)

        # pred = sess.run(y, feed_dict={
        #     x: normalized
        # })
        plt.imshow(output_data.reshape(128, 128))
        plt.show()

if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument(
        '--pb_file',
        type=str,
        default='artifacts/model.pb',
    )
    parser.add_argument(
        '--img_file',
        type=str,
        default='data/glasshair.jpg',
        help='image file as numpy format'
    )
    args, _ = parser.parse_known_args()
    main(**vars(args))

[kei9327]

Thanks @sercant

But I hav some problem...

in Shell...

Traceback (most recent call last):
  File "tf-convert-tflite.py", line 89, in <module>
    main(**vars(args))
  File "tf-convert-tflite.py", line 47, in main
    tflite_model = converter.convert()
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/lite.py", line 439, in convert
    **converter_kwargs)
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/convert.py", line 309, in toco_convert_impl
    input_data.SerializeToString())
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/convert.py", line 109, in toco_convert_protos
    (stdout, stderr))
RuntimeError: TOCO failed see console for info.
/bin/sh: toco_from_protos: command not found

None

I can't handling this shell error... So How can i do solve this problem?

[akirasosa]

Hi, I have rewrite code using PyTorch and tf-lite is TODO now.

[ldenoue]

did you notice better accuracy when using pytorch?

[sercant]

Thanks @sercant

But I hav some problem...

in Shell...

Traceback (most recent call last):
  File "tf-convert-tflite.py", line 89, in <module>
    main(**vars(args))
  File "tf-convert-tflite.py", line 47, in main
    tflite_model = converter.convert()
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/lite.py", line 439, in convert
    **converter_kwargs)
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/convert.py", line 309, in toco_convert_impl
    input_data.SerializeToString())
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/convert.py", line 109, in toco_convert_protos
    (stdout, stderr))
RuntimeError: TOCO failed see console for info.
/bin/sh: toco_from_protos: command not found

None

I can't handling this shell error... So How can i do solve this problem?

Maybe it's because of the python or tensorflow version difference. I am using Python 3.6.5 and Tensorflow 1.12.0.

[kei9327]

Thanks @sercant But I hav some problem... in Shell...

Traceback (most recent call last):
  File "tf-convert-tflite.py", line 89, in <module>
    main(**vars(args))
  File "tf-convert-tflite.py", line 47, in main
    tflite_model = converter.convert()
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/lite.py", line 439, in convert
    **converter_kwargs)
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/convert.py", line 309, in toco_convert_impl
    input_data.SerializeToString())
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/convert.py", line 109, in toco_convert_protos
    (stdout, stderr))
RuntimeError: TOCO failed see console for info.
/bin/sh: toco_from_protos: command not found

None

I can't handling this shell error... So How can i do solve this problem?

Maybe it's because of the python or tensorflow version difference. I am using Python 3.6.5 and Tensorflow 1.12.0.

If it's not an excuse, Do you have any pretrain data (TFLite)?

[sercant]

Thanks @sercant But I hav some problem... in Shell...

Traceback (most recent call last):
  File "tf-convert-tflite.py", line 89, in <module>
    main(**vars(args))
  File "tf-convert-tflite.py", line 47, in main
    tflite_model = converter.convert()
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/lite.py", line 439, in convert
    **converter_kwargs)
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/convert.py", line 309, in toco_convert_impl
    input_data.SerializeToString())
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/convert.py", line 109, in toco_convert_protos
    (stdout, stderr))
RuntimeError: TOCO failed see console for info.
/bin/sh: toco_from_protos: command not found

None

I can't handling this shell error... So How can i do solve this problem?

Maybe it's because of the python or tensorflow version difference. I am using Python 3.6.5 and Tensorflow 1.12.0.

If it's not an excuse, Do you have any pretrain data (TFLite)?

Here is the one I converted using shared pre-trained model.

[normandra]

@sercant how did the model performed on your implementation? I just tried this today and it performed really bad... not only is it slow ~1100ms it also just tries to predict a hair in the middle of the screen everytime.

[sercant]

@sercant how did the model performed on your implementation? I just tried this today and it performed really bad... not only is it slow ~1100ms it also just tries to predict a hair in the middle of the screen everytime.

Yes, it was the same case for me regarding both performance and the behavior. I don't know why it tries to find a hair in the middle of the screen all the time. Maybe the checkpoint provided in #17 was overfitted.

[chin87]

Thanks @sercant

But I hav some problem...

in Shell...

Traceback (most recent call last):
  File "tf-convert-tflite.py", line 89, in <module>
    main(**vars(args))
  File "tf-convert-tflite.py", line 47, in main
    tflite_model = converter.convert()
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/lite.py", line 439, in convert
    **converter_kwargs)
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/convert.py", line 309, in toco_convert_impl
    input_data.SerializeToString())
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/convert.py", line 109, in toco_convert_protos
    (stdout, stderr))
RuntimeError: TOCO failed see console for info.
/bin/sh: toco_from_protos: command not found

None

I can't handling this shell error... So How can i do solve this problem?

has anyone solved this error, i'm getting same error for my model

[sercant]

Thanks @sercant But I hav some problem... in Shell...

Traceback (most recent call last):
  File "tf-convert-tflite.py", line 89, in <module>
    main(**vars(args))
  File "tf-convert-tflite.py", line 47, in main
    tflite_model = converter.convert()
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/lite.py", line 439, in convert
    **converter_kwargs)
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/convert.py", line 309, in toco_convert_impl
    input_data.SerializeToString())
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/contrib/lite/python/convert.py", line 109, in toco_convert_protos
    (stdout, stderr))
RuntimeError: TOCO failed see console for info.
/bin/sh: toco_from_protos: command not found

None

I can't handling this shell error... So How can i do solve this problem?

has anyone solved this error, i'm getting same error for my model

Hey @chin87, I think this is due to a bug in Tensorflow right now. Be sure that toco_from_protos is in your PATH variable.

Also, I have an updated version of the script at my repo if you want to check it out. You will need to modify it to work with this repo though.

[charliesantos]

Hi @sercant @akirasosa sorry for bringing up this 2 year old issue. Any news on adding a script to convert to tflite? @sercant , does your script still works with this repo? Thank you in advance!