利用官方案例进行训练自己的目标检测模型

准备工作

请参考上篇

https://github.com/tensorflow/models

标注工具

https://github.com/tzutalin/labelImg

安装对应模块，调试环境即可

利用标注工具制作自己的数据集，并生成xml文件

生成csv

# -*- coding:utf-8 -*-
import os
import glob
import pandas as pd
import xml.etree.ElementTree as ET
import random
def xml_to_csv(path):
    xml_list = []
    xml_list_test = []
    rate = 0.8
    i = 0
    img_file = glob.glob(path + '/*.xml')
    print (img_file)
    random.shuffle(img_file)
    for xml_file in img_file:
        i = i + 1
        num_of_train = int(len(glob.glob(path + '/*.xml')) * rate)
        tree = ET.parse(xml_file)
        root = tree.getroot()
        if i <= num_of_train:
            for member in root.findall('object'):
                value = (
                         "dataset/pic/"+root.find('filename').text+".jpg",
                         int(root.find('size')[0].text),
                         int(root.find('size')[1].text),
                         member[1].text,
                         int(member[5][0].text),
                         int(member[5][1].text),
                         int(member[5][2].text),
                         int(member[5][3].text)
                         )
         
                print(value)
                xml_list.append(value)
        else:
            for member in root.findall('object'):
                value = (
                         "dataset/pic/"+root.find('filename').text+".jpg",
                         int(root.find('size')[0].text),
                         int(root.find('size')[1].text),
                         member[1].text,
                         int(member[5][0].text),
                         int(member[5][1].text),
                         int(member[5][2].text),
                         int(member[5][3].text)
                         )
                print(value)
                xml_list_test.append(value)
    column_name = ['filename', 'width', 'height', 'class', 'xmin', 'ymin', 'xmax', 'ymax']
    xml_df = pd.DataFrame(xml_list, columns=column_name)
    xml_df_test = pd.DataFrame(xml_list_test, columns=column_name)
    return xml_df, xml_df_test

def main():
    image_path = 'train_xml/'
    csv_save_path = 'train_labels.csv'
    csv_save_path_test = 'test_labels.csv'
    xml_df, xml_df_test = xml_to_csv(image_path)
    xml_df.to_csv(csv_save_path, index=None)
    xml_df_test.to_csv(csv_save_path_test, index=None)

main()

csv图片路径必须是绝对路径，否则报错。

生成record文件

gennrate_tfrecord_.py --csv_input=dataset/train_labels.csv --output_path=train.record --imahe_dir=dataset/pic
gennrate_tfrecord_.py --csv_input=dataset/test_labels.csv --output_path=test.record --imahe_dir=dataset/pic

gennrate_tfrecord_.py

from __future__ import division
from __future__ import print_function
from __future__ import absolute_import
 
import os
import io
import pandas as pd
try:
    import tensorflow as tf
ecaept:
	import tensorflow.compat.v1 as tf
import sys
from PIL import Image
 

from object_detection.utils import dataset_util
from collections import namedtuple, OrderedDict
 
flags = tf.app.flags
flags.DEFINE_string('csv_input', default='/home/hanqing/SSD-Tensorflow-master/VOC2019/ImageSets/Main/csv/sj_train1.csv',help='')
flags.DEFINE_string('output_path', default='/home/hanqing/SSD-Tensorflow-master/tfrecords_/sj_train.record',help='')
flags.DEFINE_string('image_dir', default='/home/hanqing/SSD-Tensorflow-master/VOC2019/JPEGImages/sj_data/',help='')
FLAGS = flags.FLAGS
 
 

def class_text_to_int(row_label):
    if row_label == "animation_person":
        return 1
    elif row_label == 'women':
        return 2
    else:
        return 0
 
 
def split(df, group):
    data = namedtuple('data', ['filename', 'object'])
    gb = df.groupby(group)
    return [data(filename, gb.get_group(x)) for filename, x in zip(gb.groups.keys(), gb.groups)]
 
 
def create_tf_example(group, path):
    with tf.gfile.GFile(os.path.join(path, '{}'.format(group.filename)), 'rb') as fid:
        encoded_jpg = fid.read()
    encoded_jpg_io = io.BytesIO(encoded_jpg)
    image = Image.open(encoded_jpg_io)
    width, height = image.size
 
    filename = group.filename.encode('utf8')
    image_format = b'jpg'
    xmins = []
    xmaxs = []
    ymins = []
    ymaxs = []
    classes_text = []
    classes = []
 
    for index, row in group.object.iterrows():
        xmins.append(row['xmin'] / width)
        xmaxs.append(row['xmax'] / width)
        ymins.append(row['ymin'] / height)
        ymaxs.append(row['ymax'] / height)
        classes_text.append(str(row['class']).encode('utf8'))
        classes.append(class_text_to_int(row['class']))
 
    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(encoded_jpg),
        '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(classes_text),
        'image/object/class/label': dataset_util.int64_list_feature(classes),
    }))
    return tf_example
 
 
def main(_):
    writer = tf.python_io.TFRecordWriter(FLAGS.output_path)
    path = os.path.join(FLAGS.image_dir)
    examples = pd.read_csv(FLAGS.csv_input)
    grouped = split(examples, 'filename')
    for group in grouped:
        tf_example = create_tf_example(group, path)
        writer.write(tf_example.SerializeToString())
    writer.close()
    output_path = os.path.join(os.getcwd(), FLAGS.output_path)
    print('Successfully created the TFRecords: {}'.format(output_path))
 
 
if __name__ == '__main__':
    tf.app.run()

切换到工程目录在进行操作

下载模型

https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf2_detection_zoo.md
https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf1_detection_zoo.md

解压放到model目录,并将三个文件放入training目录

修改label_map.pbtxt

如果找不到自己手写也可以

修改配置文件

在object_detection\samples\configs找到下载对应模型的配置文件修改信息
ssd_mobilenet_v2_coco.config




根据情况批处理这个可以改小一点

输入训练命令

model_main.py --pipeline_config_path=config/ssd_mobilenet_v2_coco.config --model_dir=training --alsologtostder

最后来了句这个，有知道怎么解决的请指教。
说明，切换文tf1.4 ctype相关错误，tf2.4 版本不对应。

全部使用tf2进行

阿来，error

解决方案

参考，https://www.youtube.com/watch?v=oqd54apcgGE