deeplabv3+使用voc2012 augmented 数据进行训练

超、凢脫俗 2024-04-17 18:40 76阅读 0赞

在原始的论文中结果使用了augmented数据的，因为改数据不是平移旋转放缩的增加方式，而是将voc2012中未标注的图片进行了标注，不过由于并不是那种很精准的标注，举个例子：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTMyNDk4NTM_size_16_color_FFFFFF_t_70][]

也没有白边，推测是利用边缘检测结合手工。

关于该数据的获取：[https://blog.csdn.net/u013249853/article/details/100136780][https_blog.csdn.net_u013249853_article_details_100136780]

# 使用方法简单到不可思议。。。 #

注意下面的代码你需要改三个地址，对应于我们上面说的[三个数据][https_blog.csdn.net_u013249853_article_details_100136780]

图片地址，分割地址（标注），list地址。

以及最终输出的文件夹叫做tfrecord放一张我的结构图片

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTMyNDk4NTM_size_16_color_FFFFFF_t_70 1][]

# Copyright 2018 The TensorFlow Authors All Rights Reserved.
    #
    # Licensed under the Apache License, Version 2.0 (the "License");
    # you may not use this file except in compliance with the License.
    # You may obtain a copy of the License at
    #
    #     http://www.apache.org/licenses/LICENSE-2.0
    #
    # Unless required by applicable law or agreed to in writing, software
    # distributed under the License is distributed on an "AS IS" BASIS,
    # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    # See the License for the specific language governing permissions and
    # limitations under the License.
    # ==============================================================================
    
    """Converts PASCAL VOC 2012 augmented data to TFRecord file format with Example protos.
    
    PASCAL VOC 2012 dataset is expected to have the following directory structure:
    
    
      ./VOCdevkit/VOC2012/ImageSets/Segmentation
    
    This script converts data into sharded data files and save at tfrecord folder.
    
    The Example proto contains the following fields:
    
      image/encoded: encoded image content.
      image/filename: image filename.
      image/format: image file format.
      image/height: image height.
      image/width: image width.
      image/channels: image channels.
      image/segmentation/class/encoded: encoded semantic segmentation content.
      image/segmentation/class/format: semantic segmentation file format.
    """
    import math
    import os.path
    import sys
    from deeplab.datasets import build_data
    import tensorflow as tf
    
    FLAGS = tf.app.flags.FLAGS
    root_dir = '/home/DATA/liutian/tmp/VOC2012_aug/benchmark_RELEASE/dataset/'
    #todo: combine aumented dataset of voc and voc into a consist tfrecord
    voc_dir = '/home'
    tf.app.flags.DEFINE_string('image_folder',
                               root_dir+'/img',
                               'Folder containing images.')
    
    tf.app.flags.DEFINE_string(
        'semantic_segmentation_folder',
        root_dir+'SegmentationClassAug/SegmentationClassAug',
        'Folder containing semantic segmentation annotations.')
    
    tf.app.flags.DEFINE_string(
        'list_folder',
        root_dir,
        'Folder containing lists for training and validation')
    
    tf.app.flags.DEFINE_string(
        'output_dir',
        root_dir+'/tfrecord',
        'Path to save converted SSTable of TensorFlow examples.')
    
    
    _NUM_SHARDS = 20
    
    
    def _convert_dataset(dataset_split):
      """Converts the specified dataset split to TFRecord format.
    
      Args:
        dataset_split: The dataset split (e.g., train, test).
    
      Raises:
        RuntimeError: If loaded image and label have different shape.
      """
      dataset = os.path.basename(dataset_split)[:-4]
      sys.stdout.write('Processing ' + dataset)
      filenames = [x.strip('\n') for x in open(dataset_split, 'r')]
      num_images = len(filenames)
      num_per_shard = int(math.ceil(num_images / float(_NUM_SHARDS)))
    
      image_reader = build_data.ImageReader('jpeg', channels=3)
      label_reader = build_data.ImageReader('png', channels=1)
    
      for shard_id in range(_NUM_SHARDS):
        output_filename = os.path.join(
            FLAGS.output_dir,
            '%s-%05d-of-%05d.tfrecord' % (dataset, shard_id, _NUM_SHARDS))
        with tf.python_io.TFRecordWriter(output_filename) as tfrecord_writer:
          start_idx = shard_id * num_per_shard
          end_idx = min((shard_id + 1) * num_per_shard, num_images)
          for i in range(start_idx, end_idx):
            sys.stdout.write('\r>> Converting image %d/%d shard %d' % (
                i + 1, len(filenames), shard_id))
            sys.stdout.flush()
            # Read the image.
            image_filename = os.path.join(
                FLAGS.image_folder, filenames[i] + '.' + FLAGS.image_format)
            image_data = tf.gfile.FastGFile(image_filename, 'rb').read()
            height, width = image_reader.read_image_dims(image_data)
            # Read the semantic segmentation annotation.
            seg_filename = os.path.join(
                FLAGS.semantic_segmentation_folder,
                filenames[i] + '.' + FLAGS.label_format)
            seg_data = tf.gfile.FastGFile(seg_filename, 'rb').read()
            seg_height, seg_width = label_reader.read_image_dims(seg_data)
            if height != seg_height or width != seg_width:
              raise RuntimeError('Shape mismatched between image and label.')
            # Convert to tf example.
            example = build_data.image_seg_to_tfexample(
                image_data, filenames[i], height, width, seg_data)
            tfrecord_writer.write(example.SerializeToString())
        sys.stdout.write('\n')
        sys.stdout.flush()
    
    
    def main(unused_argv):
      dataset_splits = tf.gfile.Glob(os.path.join(FLAGS.list_folder, 'train_aug.txt'))
      for dataset_split in dataset_splits:
        _convert_dataset(dataset_split)
    
    
    if __name__ == '__main__':
      tf.app.run()

上面的代码跑完，你应该能够得到，扩增数据的tfrecord图片。

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTMyNDk4NTM_size_16_color_FFFFFF_t_70 2][]

将上述的tfrecord文件全部复制到你的项目目录..../deeplab/datasets/pascal\_voc\_seg/tfrecord

利用下面的bash代码：

python "${WORK_DIR}"/train.py \
      --logtostderr \
      --train_split="train_aug" \
      --model_variant="xception_65" \
      --atrous_rates=6 \
      --atrous_rates=12 \
      --atrous_rates=18 \
      --output_stride=16 \
      --decoder_output_stride=4 \
      --train_crop_size="513,513" \
      --train_batch_size=15 \
      --training_number_of_steps=8000 \
      --fine_tune_batch_norm=False \
      --num_clones=5 \
      --base_learning_rate=0.007 \
      --tf_initial_checkpoint="${COCO_PRE}/x65-b2u1s2p-d48-2-3x256-sc-cr300k_init.ckpt" \
      --train_logdir="${TRAIN_LOGDIR}" \
      --dataset_dir="${PASCAL_DATASET}"\
      --initialize_last_layer=False

即可完成。

这里主要是将train\_split值设置成train\_aug即可。

通篇其实是可以不该代码的。第一个其实也可以只改bash。感兴趣自己改一下。

[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTMyNDk4NTM_size_16_color_FFFFFF_t_70]: https://image.dandelioncloud.cn/pgy_files/images/2024/04/17/95cbae78a29f45f6b221feb1cc24bb70.png
[https_blog.csdn.net_u013249853_article_details_100136780]: https://blog.csdn.net/u013249853/article/details/100136780
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTMyNDk4NTM_size_16_color_FFFFFF_t_70 1]: https://image.dandelioncloud.cn/pgy_files/images/2024/04/17/b22ef835e3fc406faf16af9f2e2f362d.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTMyNDk4NTM_size_16_color_FFFFFF_t_70 2]: https://image.dandelioncloud.cn/pgy_files/images/2024/04/17/c907a8320ca14d33b3751b67c5736ab1.png