mask_to_submission.py

#!/usr/bin/env python3

import os
import numpy as np
import matplotlib.image as mpimg
import re
import PIL

from absl import app, flags

FLAGS = flags.FLAGS

flags.DEFINE_string(
    "submission_filename",
    "submissions/test_sub_3.csv",
    "The output csv for the submission.",
)
flags.DEFINE_string(
    "base_dir", "test/predictions", "The directory with the predicted masks."
)

foreground_threshold = 0.25  # percentage of pixels of val 255 required to assign a foreground label to a patch

# assign a label to a patch
def patch_to_label(patch):
    patch = patch.astype(np.float64) / 255
    df = np.mean(patch)
    if df > foreground_threshold:
        return 1
    else:
        return 0


def mask_to_submission_strings(image_filename, mask_dir=None):
    if "DS_Store" in image_filename:
        return

    """Reads a single image and outputs the strings that should go into the submission file"""
    img_number = int(re.search(r"\d+", image_filename).group(0))
    im = PIL.Image.open(image_filename)
    im_arr = np.asarray(im)
    if len(im_arr.shape) > 2:
        # Convert to grayscale.
        im = im.convert("L")
        im_arr = np.asarray(im)

    patch_size = 16
    mask = np.zeros_like(im_arr)
    for j in range(0, im_arr.shape[1], patch_size):
        for i in range(0, im_arr.shape[0], patch_size):
            patch = im_arr[i : i + patch_size, j : j + patch_size]
            label = patch_to_label(patch)
            mask[i : i + patch_size, j : j + patch_size] = int(label * 255)
            yield ("{:03d}_{}_{},{}".format(img_number, j, i, label))

    if mask_dir:
        save_mask_as_img(
            mask, os.path.join(mask_dir, "mask_" + image_filename.split("/")[-1])
        )


def save_mask_as_img(img_arr, mask_filename):
    img = PIL.Image.fromarray(img_arr)
    os.makedirs(os.path.dirname(mask_filename), exist_ok=True)
    img.save(mask_filename)


def masks_to_submission(submission_filename, mask_dir, *image_filenames):
    """Converts images into a submission file"""
    with open(submission_filename, "w") as f:
        f.write("id,prediction\n")
        for fn in image_filenames[0:]:
            f.writelines(
                "{}\n".format(s)
                for s in mask_to_submission_strings(fn, mask_dir=mask_dir)
            )


def main(_):
    image_filenames = [
        os.path.join(FLAGS.base_dir, name) for name in os.listdir(FLAGS.base_dir)
    ]
    masks_to_submission(FLAGS.submission_filename, "", *image_filenames)


if __name__ == "__main__":
    app.run(main)