一、介绍
近两年来出现了许多优秀的分类、物体检测和分割模型,包括SSD,yolo,Mask RCNN,这些模型速度快,准确率高,被广泛应用于无人驾驶、人脸定位识别、工业缺陷检测中。工业领域,系统一般基于C/C++,而这些模型基本都都是基于python的,那么如何把python训练的模型集成到C++中呢?Opencv搭了一个很好的桥梁。opencv4.0可以支持调用Mask RCNN、Faster RCNN、SSD、YOLO等模型。这样我们可以通过线下训练好这些模型,通过opencv把模型集成到系统中。
二、opencv调用模型
关于opencv如何调用深度学习模型请参考前面两篇博文:
https://blog.csdn.net/qq_29462849/article/details/85272575
https://blog.csdn.net/qq_29462849/article/details/85262609
opencv做了一个很好的tensorflow接口,这个时候就不得不提tensorflow object detection api,里面集成了SSD、Faster RCNN、Mask RCNN等多种模型,opencv和其无缝对接。
关于tensorflow object detection api的安装和配置请参考:
https://blog.csdn.net/zong596568821xp/article/details/82015126
https://blog.csdn.net/chuquanchang1051/article/details/79804965
如果在windows下配置时,出现缺少vc++14,那么下载安装即可。windows下如果proto成功了,但是安装不上,没关系,直接把整个object_detection模块拿来用就好。
三、训练数据
由于object detection api训练时采用的是tfrecords文件,所以需要进行转换,对voc数据来说,需要把xml文件转换成csv文件,源代码如下:
import os
import glob
import pandas as pd
import xml.etree.ElementTree as ET
def xml_to_csv(path):
xml_list = []
for xml_file in glob.glob(path + '/*.xml'):
tree = ET.parse(xml_file)
root = tree.getroot()
for member in root.findall('object'):
value = (root.find('filename').text,
int(root.find('size')[0].text),
int(root.find('size')[1].text),
member[0].text,
int(member[4][0].text),
int(member[4][1].text),
int(member[4][2].text),
int(member[4][3].text)
)
xml_list.append(value)
column_name = ['filename', 'width', 'height', 'class', 'xmin', 'ymin', 'xmax', 'ymax']
xml_df = pd.DataFrame(xml_list, columns=column_name)
return xml_df
def main():
image_path = os.path.join('./train_data/tank/test/', 'Annotations')
xml_df = xml_to_csv(image_path)
xml_df.to_csv('./train_data/tank/test/csv/test.csv', index=None)
print('Successfully converted xml to csv.')
main()
当把xml文件转换成csv文件后,需要对csv文件进行转换,生成最终得tfrecords文件,转换代码如下所示:
"""
Usage:
# From tensorflow/models/
# Create train data:
python generate_tfrecord.py --csv_input=data/train_labels.csv --output_path=train.record
# Create test data:
python generate_tfrecord.py --csv_input=data/test_labels.csv --output_path=test.record
"""
from __future__ import division
from __future__ import print_function
from __future__ import absolute_import
import os
import io
import pandas as pd
import tensorflow as tf
import sys
from PIL import Image
# sys.path.append("F:\setup\tf\models-master\research\object_detection")
from object_detection.utils import dataset_util
from collections import namedtuple, OrderedDict
flags = tf.app.flags
flags.DEFINE_string('csv_input', default='./train_data/tank/train/csv/train.csv',help='')
flags.DEFINE_string('output_path', default='./train_data/tank/train/tf_record/train.record',help='')
flags.DEFINE_string('image_dir', default='./train_data/tank/train/JPEGImages/',help='')
FLAGS = flags.FLAGS
# TO-DO replace this with label map
def class_text_to_int(row_label):
if row_label == 'tank':
return 1
elif row_label ==