tensorflow object detection api v1训练自己的数据集详细教程

1. 安装

• 拉取镜像

  docker pull nvcr.io/nvidia/tensorflow:21.02-tf1-py3
  

• 启动容器

  container_name=tf21.02
  docker run -itd –name {container_name} -v /workspace:/workspace nvcr.io/nvidia/tensorflow:21.02-tf1-py3 /bin/bash
  

• 进入容器

  docker exec -it {container_name} /bin/bash
  

• 安装依赖

  apt-get update && apt-get install -y \
      git \
      gpg-agent \
      python3-cairocffi \
      protobuf-compiler \
      python3-pil \
      python3-lxml \
      python3-tk \
      wget

• 安装tf object detection api v2

  cd /opt
  git clone https://github.com/tensorflow/models.git
  cp models/ /root/models
  cd /root/models/research
  protoc object_detection/protos/*.proto --python_out=.  # 编译proto文件
  cp object_detection/packages/tf2/setup.py .  
  pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple  # 换成清华源
  python -m pip install .
  python object_detection/builders/model_builder_tf2_test.py   # 测试是否安装成功
  

2. 制作自定义数据集

• 首先将数据集格式制作成VOC格式

  VOCdevkit
  -- VOC2007
     -- Annotations
        -- *.xml
     -- JPEGImages
        -- *.jpg
     -- ImageSets
        -- Main
           -- train.txt
           -- test.txt  
  

• 过滤掉不是JPEG编码的图片

  import PIL.Image
  import os
  imgs_folder = r'/workspace/VOCdevkit/VOC2007/JPEGImages'   # jpg 存放目录
  annos_folder = r'/workspace/VOCdevkit/VOC2007/Annotations'  # xml 存放目录
  imgs_filename = os.listdir(imgs_folder)
  
  for filename in imgs_filename:
      img_fullpath = os.path.join(img_folder, filename)
      image = PIL.Image.open(img_fullpath)
  
      if image.format != 'JPEG':
          # print(img_fullpath)
          os.remove(img_fullpath)
          anno_filename = filename[:-4] + '.xml'
          anno_fullpath = os.path.join(annos_folder, anno_filename)
          os.remove(anno_fullpath)
  
  

• 转tfrecord

  • 在/root/models/research/object_detection 目录下创建create_pascal_tf_record.py 脚本

    vim create_pascal_tf_record.py
    

  • 将以下代码复制到create_pascal_tf_record.py中

    # Copyright 2017 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.
    # ==============================================================================
    
    r"""Convert raw PASCAL dataset to TFRecord for object_detection.
    
    Example usage:
        python object_detection/dataset_tools/create_pascal_tf_record.py \
            --data_dir=/home/user/VOCdevkit \
            --year=VOC2012 \
            --output_path=/home/user/pascal.record
    """
    from __future__ import absolute_import
    from __future__ import division
    from __future__ import print_function
    
    import hashlib
    import io
    import logging
    import os
    
    from lxml import etree
    import PIL.Image
    import tensorflow.compat.v1 as tf
    
    from object_detection.utils import dataset_util
    from object_detection.utils import label_map_util
    
    
    flags = tf.app.flags
    flags.DEFINE_string('data_dir', '', 'Root directory to raw PASCAL VOC dataset.')
    flags.DEFINE_string('set', 'train', 'Convert training set, validation set or '
                        'merged set.')
    flags.DEFINE_string('annotations_dir', 'Annotations',
                        '(Relative) path to annotations directory.')
    flags.DEFINE_string('year', 'VOC2007', 'Desired challenge year.')
    flags.DEFINE_string('output_path', '', 'Path to output TFRecord')
    flags.DEFINE_string('label_map_path', 'data/pascal_label_map.pbtxt',
                        'Path to label map proto')
    flags.DEFINE_boolean('ignore_difficult_instances', False, 'Whether to ignore '
                         'difficult instances')
    FLAGS = flags.FLAGS
    
    SETS = ['train', 'val', 'trainval', 'test']
    YEARS = ['VOC2007', 'VOC2012', 'merged']
    
    
    def dict_to_tf_example(data,
                           dataset_directory,
                           label_map_dict,
                           ignore_difficult_instances=False,
                           image_subdirectory='JPEGImages'):
      """Convert XML derived dict to tf.Example proto.
    
      Notice that this function normalizes the bounding box coordinates provided
      by the raw data.
    
      Args:
        data: dict holding PASCAL XML fields for a single image (obtained by
          running dataset_util.recursive_parse_xml_to_dict)
        dataset_directory: Path to root directory holding PASCAL dataset
        label_map_dict: A map from string label names to integers ids.
        ignore_difficult_instances: Whether to skip difficult instances in the
          dataset  (default: False).
        image_subdirectory: String specifying subdirectory within the
          PASCAL dataset directory holding the actual image data.
    
      Returns:
        example: The converted tf.Example.
    
      Raises:
        ValueError: if the image pointed to by data['filename'] is not a valid JPEG
      """
      img_path = os.path.join('VOC2007','JPEGImages',  data['filename'])
      full_path = os.path.join(dataset_directory, img_path)
      with tf.gfile.GFile(full_path, 'rb') as fid:
        encoded_jpg = fid.read()
      encoded_jpg_io = io.BytesIO(encoded_jpg)
      image = PIL.Image.open(encoded_jpg_io)
    
      if image.format != 'JPEG':
        print(full_path)
        raise ValueError('Image format not JPEG')
      key = hashlib.sha256(encoded_jpg).hexdigest()
    
      width = int(data['size']['width'])
      height = int(data['size']['height'])
    
      xmin = []
      ymin = []
      xmax = []
      ymax = []
      classes = []
      classes_text = []
      truncated = []
      poses = []
      difficult_obj = []
      if 'object' in data:
        for obj in data['object']:
          difficult = bool(int(obj['difficult']))
          if ignore_difficult_instances and difficult:
            continue
    
          difficult_obj.append(int(difficult))
    
          xmin.append(float(obj['bndbox']['xmin']) / width)
          ymin.append(float(obj['bndbox']['ymin']) / height)
          xmax.append(float(obj['bndbox']['xmax']) / width)
          ymax.append(float(obj['bndbox']['ymax']) / height)
          classes_text.append(obj['name'].encode('utf8'))
          classes.append(label_map_dict[obj['name']])
          truncated.append(int(obj['truncated']))
          poses.append(obj['pose'].encode('utf8'))
    
      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(
              data['filename'].encode('utf8')),
          'image/source_id': dataset_util.bytes_feature(
              data['filename'].encode('utf8')),
          'image/key/sha256': dataset_util.bytes_feature(key.encode('utf8')),
          'image/encoded': dataset_util.bytes_feature(encoded_jpg),
          'image/format': dataset_util.bytes_feature('jpeg'.encode('utf8')),
          'image/object/bbox/xmin': dataset_util.float_list_feature(xmin),
          'image/object/bbox/xmax': dataset_util.float_list_feature(xmax),
          'image/object/bbox/ymin': dataset_util.float_list_feature(ymin),
          'image/object/bbox/ymax': dataset_util.float_list_feature(ymax),
          'image/object/class/text': dataset_util.bytes_list_feature(classes_text),
          'image/object/class/label': dataset_util.int64_list_feature(classes),
          'image/object/difficult': dataset_util.int64_list_feature(difficult_obj),
          'image/object/truncated': dataset_util.int64_list_feature(truncated),
          'image/object/view': dataset_util.bytes_list_feature(poses),
      }))
      return example
    
    
    def main(_):
      if FLAGS.set not in SETS:
        raise ValueError('set must be in : {}'.format(SETS))
      if FLAGS.year not in YEARS:
        raise ValueError('year must be in : {}'.format(YEARS))
    
      data_dir = FLAGS.data_dir
      years = ['VOC2007', 'VOC2012']
      if FLAGS.year != 'merged':
        years = [FLAGS.year]
    
      writer = tf.python_io.TFRecordWriter(FLAGS.output_path)
    
      label_map_dict = label_map_util.get_label_map_dict(FLAGS.label_map_path)
    
      for year in years:
        logging.info('Reading from PASCAL %s dataset.', year)
        examples_path = os.path.join(data_dir, year, 'ImageSets', 'Main',
                                     FLAGS.set + '.txt')
        annotations_dir = os.path.join(data_dir, year, FLAGS.annotations_dir)
        examples_list = dataset_util.read_examples_list(examples_path)
        for idx, example in enumerate(examples_list):
          if idx % 100 == 0:
            logging.info('On image %d of %d', idx, len(examples_list))
          path = os.path.join(annotations_dir, example + '.xml')
          with tf.gfile.GFile(path, 'r') as fid:
            xml_str = fid.read()
          xml = etree.fromstring(xml_str.encode('utf-8'))
          data = dataset_util.recursive_parse_xml_to_dict(xml)['annotation']
    
          tf_example = dict_to_tf_example(data, FLAGS.data_dir, label_map_dict,
                                          FLAGS.ignore_difficult_instances)
          writer.write(tf_example.SerializeToString())
    
      writer.close()
    
    
    if __name__ == '__main__':
      tf.app.run()
    

  • 制作label_map.pbtxt

    vim /workspace/label_map.pbtxt
    
    # 将以下内容复制进去，类别id从1开始，不包括背景。
    item {
      id: 1
      name: 'car'
    }
    

  • 生成train.tfrecord, test.tfrecord

    
    
    dataset_dir = '{your_path}/VOCdevkit'  # eg: /Munyun/VOCdevkit
    # 生成train.tfrecord
    python create_pascal_tf_record.py \
           --data_dir={dataset_dir} \
           --set=train \
           --year=VOC2007 \
           --output_path={your_path}/train.tfrecord \
           --label_map_path={your_path}/label_map.txt 
           
    # 生成test.tfrecord
    python create_pascal_tf_record.py \
           --data_dir={dataset_dir} \
           --set=test \
           --year=VOC2007 \
           --output_path={your_path}/test.tfrecord \
           --label_map_path={your_path}/label_map.txt
    

训练：
python create_anfang_tfrecord.py --data_dir=/home/Mycode/VOCdevkit_anfang --set=train --year=VOC2007 --output_path=/home/Mycode/train.tfrecord --label_map_path=anfang_label_map.pbtxt

测试：
python create_anfang_tfrecord.py --data_dir=/home/Mycode/VOCdevkit_anfang --set=test --year=VOC2007 --output_path=/home/Mycode/test.tfrecord --label_map_path=anfang_label_map.pbtxt

3. 训练

修改model config

cp object_detection/configs/tf2/faster_rcnn_resnet101_v1_800x1333_coco17_gpu.config \
/Muyunfaster_rcnn_resnet101_v1_800x1333_coco17_gpu.config

（1）修改num_classes
    num_classes=4 （自己数据集的类别数）
（2）注释掉fine_tune_checkpoint
（3）修改fine_tune_checkpoint_type
    fine_tune_checkpoint_type="detection"  # detection 检测、classifcial  分类
（4）修改train_input_reader
    train_input_reader: {
      # 此处修改为{your_path}/label_map.txt
      label_map_path: "PATH_TO_BE_CONFIGURED/label_map.txt"
      tf_record_input_reader {
        # 此处修改为{your_path}/train.tfrecord
        input_path: "PATH_TO_BE_CONFIGURED/train.tfrecord"
      } 
    }
（5）修改eval_input_reader
    eval_input_reader: {
      # 此处修改为{your_path}/label_map.txt
      label_map_path: "PATH_TO_BE_CONFIGURED/label_map.txt"
      shuffle: false
      num_epochs: 1
      tf_record_input_reader {
        # 此处修改为{your_path}/test.tfrecord
        input_path: "PATH_TO_BE_CONFIGURED/val2017-?????-of-00032.tfrecord"
      }
    }

• 执行训练命令

  # From the tensorflow/models/research/ directory
  PIPELINE_CONFIG_PATH={path to pipeline config file}
  MODEL_DIR={path to model directory}eg:/home/mycode/logs
  NUM_TRAIN_STEPS=50000
  SAMPLE_1_OF_N_EVAL_EXAMPLES=1
  python object_detection/model_main.py \
      --pipeline_config_path=${PIPELINE_CONFIG_PATH} \
      --model_dir=${MODEL_DIR} \
      --num_train_steps=${NUM_TRAIN_STEPS} \
      --sample_1_of_n_eval_examples=${SAMPLE_1_OF_N_EVAL_EXAMPLES} \
      --alsologtostderr
      --checkpoint_dir=gs://${MODEL_DIR} # 评估加上即可 eg:/home/Mycode/logs/model.ckpt-3755
      (其中包含3个文件：model.ckpt-3755.data-00000-of-00001，model.ckpt-3755.index， model.ckpt-3755.meta)
  

5. Dockerfile

• 拉镜像
• 上传镜像

容器打包成镜像
docker commit [OPTIONS] CONTAINER [REPOSITORY[:TAG]]
-a :提交的镜像作者；

-c :使用Dockerfile指令来创建镜像；

-m :提交时的说明文字；

-p :在commit时，将容器暂停。

docker commit -a "Muyun" -m "tf目标检测api tf2.5" -p od-tf2.4 songguangfu/tf-od-api:od-tf2.5


首先登陆docker hub
然后创建一个仓库   用户名/仓库名    这里会得到一个uuid形式的密码
然后在服务器上先登陆服务器  docker login -u 用户名
然后输入密码

重命名TAG
docker tag 原io:原Tag 新IO:新Tag

docker push 用户名/仓库名:Tag

FROM nvcr.io/nvidia/tensorflow:21.02-tf2-py3
  
RUN apt-get update && apt-get install -y \
    git \
    gpg-agent \
    python3-cairocffi \
    protobuf-compiler \
    python3-pil \
    python3-lxml \
    python3-tk \
    wget
COPY models/ /root/models
WORKDIR /root/models/research
RUN  protoc object_detection/protos/*.proto --python_out=.
RUN  cp object_detection/packages/tf2/setup.py .
RUN  pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple
# RUN  python setup.py install  # 用这个是更好一些
RUN  python -m pip install .
RUN  export CUDA_VISIBLE_DEVICES=0
RUN  python object_detection/builders/model_builder_tf2_test.py

https://github.com/tensorflow/models/blob/master/research/object_detection/export_inference_graph.py
https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/oid_inference_and_evaluation.md