Last active
March 14, 2018 08:25
-
-
Save TopHatCroat/38302cc7fe31ab994d688d3623d2c5df to your computer and use it in GitHub Desktop.
tfrecord.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| import os | |
| import io | |
| import re | |
| import random | |
| import tensorflow as tf | |
| from PIL import Image | |
| from object_detection.utils import dataset_util | |
| flags = tf.app.flags | |
| flags.DEFINE_string('input_path', '', 'Path to images input directory') | |
| flags.DEFINE_string('classes', '', 'Space separated list of classes') | |
| flags.DEFINE_string('output_path', '', 'Path to output TFRecord') | |
| FLAGS = flags.FLAGS | |
| VALID_EXTENSIONS = [".jpg", ".jpeg", ".jpe", ".png"] | |
| label_map = {} | |
| testing_validation_factor = 0.8 | |
| def load_images_paths(dir): | |
| result = [] | |
| for f in os.listdir(dir): | |
| ext = os.path.splitext(f)[1] | |
| if ext.lower() not in VALID_EXTENSIONS: | |
| continue | |
| result.append(os.path.join(dir, f)) | |
| return result | |
| def create_tf_example(img_path, class_text, class_id): | |
| with tf.gfile.GFile(img_path, 'rb') as fid: | |
| image_file = fid.read() | |
| encoded_image = io.BytesIO(image_file) | |
| image = Image.open(encoded_image) | |
| width, height = image.size | |
| image_name_and_ext = img_path.split("/")[-1].split(".") | |
| filename = image_name_and_ext[0].encode('utf8') | |
| # image_format = b'jpg' | |
| image_format = image_name_and_ext[1].encode('utf8') | |
| xmins = [0] | |
| xmaxs = [1] | |
| ymins = [0] | |
| ymaxs = [1] | |
| tf_example = tf.train.Example(features=tf.train.Features(feature={ | |
| 'image/height': dataset_util.int64_feature(height), | |
| 'image/width': dataset_util.int64_feature(width), | |
| 'image/filename': dataset_util.bytes_feature(filename), | |
| 'image/source_id': dataset_util.bytes_feature(filename), | |
| 'image/encoded': dataset_util.bytes_feature(image_file), | |
| 'image/format': dataset_util.bytes_feature(image_format), | |
| 'image/object/bbox/xmin': dataset_util.float_list_feature(xmins), | |
| 'image/object/bbox/xmax': dataset_util.float_list_feature(xmaxs), | |
| 'image/object/bbox/ymin': dataset_util.float_list_feature(ymins), | |
| 'image/object/bbox/ymax': dataset_util.float_list_feature(ymaxs), | |
| 'image/object/class/text': dataset_util.bytes_list_feature([class_text.encode('utf8')]), | |
| 'image/object/class/label': dataset_util.int64_list_feature([int(class_id)]), | |
| })) | |
| return tf_example | |
| def load_label_map(): | |
| with open("./label_map.pbtxt", "r") as file: | |
| file_content = file.read().replace("\'", "").split(sep=None) | |
| file_content = "".join(file_content) | |
| id_re = re.compile("id:(\d+)") | |
| name_re = re.compile("name:(\w+)") | |
| for i in file_content.split("item")[1:]: | |
| item_id = id_re.search(i).group(1) | |
| item_name = name_re.search(i).group(1) | |
| label_map.update({item_name: item_id}) | |
| def id_for_class_name(c): | |
| if len(label_map) == 0: | |
| load_label_map() | |
| return label_map.get(c) | |
| def main(_): | |
| writer = tf.python_io.TFRecordWriter(FLAGS.output_path) | |
| validation_writer = tf.python_io.TFRecordWriter("valid_" + FLAGS.output_path) | |
| classes = ["table", "chair"] | |
| for c in classes: | |
| images = load_images_paths(os.path.join(FLAGS.input_path, c)) | |
| random.shuffle(images) | |
| for index, i in enumerate(images): | |
| example = create_tf_example(i, c, id_for_class_name(c)) | |
| if index / len(images) > testing_validation_factor: | |
| validation_writer.write(example.SerializeToString()) | |
| else: | |
| writer.write(example.SerializeToString()) | |
| writer.close() | |
| validation_writer.close() | |
| if __name__ == '__main__': | |
| tf.app.run() |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment