onnxruntime/tools/python/onnx_test_data_utils.py

import argparse
import glob
import os
import sys

import numpy as np
import onnx
from onnx import numpy_helper


def read_tensorproto_pb_file(filename):
    """Return tuple of tensor name and numpy.ndarray of the data from a pb file containing a TensorProto."""

    tensor = onnx.load_tensor(filename)
    np_array = numpy_helper.to_array(tensor)
    return tensor.name, np_array


def dump_tensorproto_pb_file(filename):
    """Dump the data from a pb file containing a TensorProto."""

    name, data = read_tensorproto_pb_file(filename)
    print("Name: {}".format(name))
    print("Shape: {}".format(data.shape))
    print(data)


def dump_pb(dir_or_filename):
    """Dump the data from either a single .pb file, or all .pb files in a directory.
    All files must contain a serialized TensorProto."""

    if os.path.isdir(dir_or_filename):
        for f in glob.glob(os.path.join(dir_or_filename, "*.pb")):
            print(f)
            dump_tensorproto_pb_file(f)
    else:
        dump_tensorproto_pb_file(dir_or_filename)


def numpy_to_pb(name, np_data, out_filename):
    """Convert numpy data to a protobuf file."""

    tensor = numpy_helper.from_array(np_data, name)
    onnx.save_tensor(tensor, out_filename)


def image_to_numpy(filename, shape, channels_last, add_batch_dim):
    """Convert an image file into a numpy array."""

    import PIL.Image  # from 'Pillow' package

    img = PIL.Image.open(filename)
    if shape:
        w, h = img.size
        new_w = shape[1]
        new_h = shape[0]

        # use the dimension that needs to shrink the least to resize to an image where that dimension matches the
        # target size.
        w_ratio = new_w / w
        h_ratio = new_h / h
        ratio = w_ratio if w_ratio > h_ratio else h_ratio
        interim_w = int(w * ratio)
        interim_h = int(h * ratio)
        img = img.resize((interim_w, interim_h), PIL.Image.ANTIALIAS)

        # center crop to the final target size
        left = (interim_w - new_w) / 2
        top = (interim_h - new_h) / 2
        right = (interim_w + new_w) / 2
        bottom = (interim_h + new_h) / 2
        img = img.crop((left, top, right, bottom))

    img_as_np = np.array(img).astype(np.float32)
    if not channels_last:
        # HWC to CHW
        img_as_np = np.transpose(img_as_np, (2, 0, 1))

    if add_batch_dim:
        # to NCHW or NHWC
        img_as_np = np.expand_dims(img_as_np, axis=0)

    return img_as_np


def create_random_data(shape, type, minvalue, maxvalue, seed):
    nptype = np.dtype(type)
    np.random.seed(seed)
    return ((maxvalue - minvalue) * np.random.sample(shape) + minvalue).astype(nptype)


def update_name_in_pb(filename, name, output_filename):
    """Update the name of the tensor in the pb file."""

    tensor = onnx.load_tensor(filename)
    tensor.name = name

    if not output_filename:
        output_filename = filename

    onnx.save_tensor(tensor, output_filename)


def get_arg_parser():
    parser = argparse.ArgumentParser(
        description="""
        Utilities for working with the input/output protobuf files used by the ONNX test cases and onnx_test_runner.
        These are expected to only contain a serialized TensorProto.

        dump_pb: Dumps the TensorProto data from an individual pb file, or all pb files in a directory.
        numpy_to_pb: Convert numpy array saved to a file with numpy.save() to a TensorProto, and serialize to a pb file.
        image_to_pb: Convert data from an image file into a TensorProto, and serialize to a pb file.
        random_to_pb: Create a TensorProto with random data, and serialize to a pb file.
        update_name_in_pb: Update the TensorProto.name value in a pb file.
                           Updates the input file unless --output <filename> is specified.
        """,
        formatter_class=argparse.RawDescriptionHelpFormatter,
    )

    parser.add_argument(
        "--action",
        help="Action to perform",
        choices=["dump_pb", "numpy_to_pb", "image_to_pb", "random_to_pb", "update_name_in_pb"],
        required=True,
    )

    parser.add_argument("--input", help="The input filename or directory name")
    parser.add_argument("--name", help="The value to set TensorProto.name to if creating/updating one.")
    parser.add_argument("--output", help="Filename to serialize the TensorProto to.")

    image_to_pb_group = parser.add_argument_group("image_to_pb", "image_to_pb specific options")
    image_to_pb_group.add_argument(
        "--resize",
        default=None,
        type=lambda s: [int(item) for item in s.split(",")],
        help="Provide the height and width to resize to as comma separated values."
        " e.g. --shape 200,300 will resize to height 200 and width 300.",
    )
    image_to_pb_group.add_argument(
        "--channels_last", action="store_true", help="Transpose image from channels first to channels last."
    )
    image_to_pb_group.add_argument(
        "--add_batch_dim",
        action="store_true",
        help="Prepend a batch dimension with value of 1 to the shape. " "i.e. convert from CHW to NCHW",
    )

    random_to_pb_group = parser.add_argument_group("random_to_pb", "random_to_pb specific options")
    random_to_pb_group.add_argument(
        "--shape",
        type=lambda s: [int(item) for item in s.split(",")],
        help="Provide the shape as comma separated values e.g. --shape 200,200",
    )
    random_to_pb_group.add_argument(
        "--datatype",
        help="numpy dtype value for the data type. e.g. f4=float32, i8=int64. "
        "See: https://docs.scipy.org/doc/numpy/reference/arrays.dtypes.html",
    )
    random_to_pb_group.add_argument(
        "--min_value", default=0, type=int, help="Limit the generated values to this minimum."
    )
    random_to_pb_group.add_argument(
        "--max_value", default=1, type=int, help="Limit the generated values to this maximum."
    )
    random_to_pb_group.add_argument(
        "--seed", default=None, type=int, help="seed to use for the random values so they're deterministic."
    )

    return parser


if __name__ == "__main__":
    arg_parser = get_arg_parser()
    args = arg_parser.parse_args()

    if args.action == "dump_pb":
        if not args.input:
            print("Missing argument. Need input to be specified.", file=sys.stderr)
            sys.exit(-1)
        np.set_printoptions(precision=10)
        dump_pb(args.input)
    elif args.action == "numpy_to_pb":
        if not args.input or not args.output or not args.name:
            print("Missing argument. Need input, output and name to be specified.", file=sys.stderr)
            sys.exit(-1)
        # read data saved with numpy
        data = np.load(args.input)
        numpy_to_pb(args.name, data, args.output)
    elif args.action == "image_to_pb":
        if not args.input or not args.output or not args.name:
            print("Missing argument. Need input, output, name to be specified.", file=sys.stderr)
            sys.exit(-1)

        img_np = image_to_numpy(args.input, args.resize, args.channels_last, args.add_batch_dim)
        numpy_to_pb(args.name, img_np, args.output)
    elif args.action == "random_to_pb":
        if not args.output or not args.shape or not args.datatype or not args.name:
            print("Missing argument. Need output, shape, datatype and name to be specified.", file=sys.stderr)
            sys.exit(-1)

        data = create_random_data(args.shape, args.datatype, args.min_value, args.max_value, args.seed)
        numpy_to_pb(args.name, data, args.output)
    elif args.action == "update_name_in_pb":
        if not args.input or not args.name:
            print("Missing argument. Need input and name to be specified.", file=sys.stderr)
            sys.exit(-1)

        update_name_in_pb(args.input, args.name, args.output)
    else:
        print("Unknown action.", file=sys.stderr)
        arg_parser.print_help(sys.stderr)