keras2onnx

	Linux	Windows
keras.io
tf.keras

Introduction

The keras2onnx model converter enables users to convert Keras models into the ONNX model format. Initially, the Keras converter was developed in the project onnxmltools. keras2onnx converter development was moved into an independent repository to support more kinds of Keras models and reduce the complexity of mixing multiple converters.

All Keras layers have been supported for conversion using keras2onnx since ONNX opset 7. Please refer to the Keras documentation for details on Keras layers. The keras2onnx converter also supports the lambda/custom layer by working with the tf2onnx converter which is embedded directly into the source tree to avoid version conflicts and installation complexity.

Windows Machine Learning (WinML) users can use WinMLTools to convert their Keras models to the ONNX format. If you want to use the keras2onnx converter, please refer to the WinML Release Notes to identify the corresponding ONNX opset for your WinML version.

keras2onnx has been tested on Python 3.5, 3.6, and 3.7, with tensorflow 1.x (CI build). It does not support Python 2.x.

Notes

tf.keras v.s. keras.io

Both Keras model types are now supported in the keras2onnx converter. If the user's Keras package was installed from Keras.io, the converter converts the model as it was created by the keras.io package. Otherwise, it will convert it through tf.keras.

If you want to override this behaviour, please specify the environment variable TF_KERAS=1 before invoking the converter python API.

Usage

Before running the converter, please notice that tensorflow has to be installed in your python environment, you can choose tensorflow package(CPU version) or tensorflow-gpu(GPU version)

Validated pre-trained Keras models

We converted successfully for all the keras application models, and several other pretrained models. See below:

The following models need customed conversion, see the instruction column.

Scripts

It will be useful to convert the models from Keras to ONNX from a python script. You can use the following API:

import keras2onnx
keras2onnx.convert_keras(model, name=None, doc_string='', target_opset=None, channel_first_inputs=None):
    # type: (keras.Model, str, str, int, []) -> onnx.ModelProto
    """
    :param model: keras model
    :param name: the converted onnx model internal name
    :param doc_string:
    :param target_opset:
    :param channel_first_inputs: A list of channel first input.
    :return:
    """

Use the following script to convert keras application models to onnx, and then perform inference:

import numpy as np
from keras.preprocessing import image
from keras.applications.resnet50 import preprocess_input
import keras2onnx
import onnxruntime

# image preprocessing
img_path = 'street.jpg'   # make sure the image is in img_path
img_size = 224
img = image.load_img(img_path, target_size=(img_size, img_size))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)

# load keras model
from keras.applications.resnet50 import ResNet50
model = ResNet50(include_top=True, weights='imagenet')

# convert to onnx model
onnx_model = keras2onnx.convert_keras(model, model.name)

# runtime prediction
content = onnx_model.SerializeToString()
sess = onnxruntime.InferenceSession(content)
x = x if isinstance(x, list) else [x]
feed = dict([(input.name, x[n]) for n, input in enumerate(sess.get_inputs())])
pred_onnx = sess.run(None, feed)

An alternative way to load onnx model to runtime session is to save the model first:

import onnx
temp_model_file = 'model.onnx'
onnx.save_model(onnx_model, temp_model_file)
sess = onnxruntime.InferenceSession(temp_model_file)

Contribute

We welcome contributions in the form of feedback, ideas, or code.

License

MIT License