前言：
系统环境为：win10，本次用的是tensorflow1.12.0
完整目录为：
python人工智能 利用Tensorflow object detection API 搭建物体识别模型（一）CPU版
python人工智能 利用Tensorflow object detection API 搭建物体识别模型（二）CPU版
python人工智能 利用Tensorflow object detection API 搭建物体识别模型（三）CPU版
上篇我们已经手动标记好了鱼，我们开始训练自己的模型
一、开始训练
进入如下目录D:\ai\objDetect3\mubiaojiance，运行代码
python D:/ai/objDetect3/models/research/object_detection/model_main.py --pipeline_config_path=training/ssdlite_mobilenet_v2_coco.config --model_dir=training --alsologtostderr
接着会输出一大堆warning，不用管，继续等，直到输出下面红色框那一行，说明运行成功了，已经开始在训练了。
我们可以看到D:\ai\objDetect3\mubiaojiance\training生成了一些数据：model.ckpt-0.meta
多等一会，至少等它训练10分钟以上，我们看下这个目录：
它每隔一段时间就会备份一个模型，我们最终要的是model.ckpt-xxx数字最大的那个文件。目前等待即可，也可以操作下一步，用图表的方式查看训练进度。
二、监控训练进度
模型训练稳定地进行后，我们进入如下目录，打开一个新的conda环境，运行命令
cd /d D:\ai\objDetect3\mubiaojiance tensorboard --logdir=training
可以看到监控的本地地址：http://localhost:6006，用浏览器访问一下
监控界面如下：
三、测试一下我们训练好的模型
模型训练的越久，准确度越高。
（1）导出最新的那个训练好的模型
我们看到574这个数字最大，就取这个替换到下面代码中--trained_checkpoint_prefix字段，然后在D:\ai\objDetect3\mubiaojiance目录下执行：
cd /d D:\ai\objDetect3\mubiaojiance set PYTHONPATH=D:\ai\objDetect3\models;D:\ai\objDetect3\models\research;D:\ai\objDetect3\models\research\slim python D:/ai/objDetect3/models/research/object_detection/export_inference_graph.py --input_type=image_tensor --pipeline_config_path=training/ssdlite_mobilenet_v2_coco.config --trained_checkpoint_prefix=training/model.ckpt-574 --output_directory=fish_inference_graph
它自动生成了fish_inference_graph目录和模型文件。导出成功
（2）代码测试
我们直接用D:\ai\objDetect3\mubiaojiance\n01440764作为测试图片
我们在D:\ai\objDetect3\mubiaojiance打开jupyter notebook
cd /d D:\ai\objDetect3\mubiaojiance jupyter notebook
再打开的网页中，新建一个python文件，命名为：fish_detection
同时输入如下代码：
#导入代码库
import numpy as np
import os
import matplotlib.pyplot as plt
from PIL import Image
import tensorflow as tf
from object_detection.utils import label_map_util
from object_detection.utils import visualization_utils as vis_util
#以便在网页中绘制图像
%matplotlib inline
#加载封装好的模型
MODEL_NAME = 'fish_inference_graph'
PATH_TO_CKPT = MODEL_NAME + '/frozen_inference_graph.pb'
detection_graph = tf.Graph()
with detection_graph.as_default():
od_graph_def = tf.GraphDef()
with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid:
serialized_graph = fid.read()
od_graph_def.ParseFromString(serialized_graph)
tf.import_graph_def(od_graph_def, name='')
#加载标签映射文件，label_map中文叫做标签映射
NUM_CLASSES = 1
PATH_TO_LABELS = 'training/my_label_map.pbtxt'
label_map = label_map_util.load_labelmap(PATH_TO_LABELS)
categories = label_map_util.convert_label_map_to_categories(label_map, max_num_classes=NUM_CLASSES, use_display_name=True)
category_index = label_map_util.create_category_index(categories)
#图片数据转换为numpy的ndarray对象
def load_image_into_numpy_array(image):
(im_width, im_height) = image.size
return np.array(image.getdata()).reshape(
(im_height, im_width, 3)).astype(np.uint8)
#从文件夹中随机选10张图片
import random
dir_path = 'n01440764'
imageName_list = os.listdir(dir_path)
imagePath_list = [os.path.join(dir_path, imageName) for imageName in imageName_list]
selected_imagePath_list = random.sample(imagePath_list, 10)
selected_imagePath_list
#用10张图片测试模型的目标检测效果
with detection_graph.as_default():
with tf.Session(graph=detection_graph) as sess:
# Definite input and output Tensors for detection_graph
image_tensor = detection_graph.get_tensor_by_name('image_tensor:0')
# Each box represents a part of the image where a particular object was detected.
detection_boxes = detection_graph.get_tensor_by_name('detection_boxes:0')
# Each score represent how level of confidence for each of the objects.
# Score is shown on the result image, together with the class label.
detection_scores = detection_graph.get_tensor_by_name('detection_scores:0')
detection_classes = detection_graph.get_tensor_by_name('detection_classes:0')
num_detections = detection_graph.get_tensor_by_name('num_detections:0')
for imagePath in selected_imagePath_list:
image = Image.open(imagePath)
# the array based representation of the image will be used later in order to prepare the
# result image with boxes and labels on it.
image_np = load_image_into_numpy_array(image)
# Expand dimensions since the model expects images to have shape: [1, None, None, 3]
image_np_expanded = np.expand_dims(image_np, axis=0)
# Actual detection.
(boxes, scores, classes, num) = sess.run(
[detection_boxes, detection_scores, detection_classes, num_detections],
feed_dict={image_tensor: image_np_expanded})
# Visualization of the results of a detection.
vis_util.visualize_boxes_and_labels_on_image_array(
image_np,
np.squeeze(boxes),
np.squeeze(classes).astype(np.int32),
np.squeeze(scores),
category_index,
use_normalized_coordinates=True,
line_thickness=8)
final_score = np.squeeze(scores)
count = 0
retData = []
for i in range(100):
if scores is None or final_score[i] > 0.5:
count = count + 1
(im_width, im_height) = image.size
for i in range(count):
#print(boxes[0][i])
y_min = boxes[0][i][0]*im_height
x_min = boxes[0][i][1]*im_width
y_max = boxes[0][i][2]*im_height
x_max = boxes[0][i][3]*im_width
retData.append({"imagePath":imagePath, "index":i, "left":int(x_min), "top":int(y_min), "width":int(x_max-x_min), "height":int(y_max-y_min), "score":final_score[i]})
print(retData)
plt.figure(figsize=(12,8))
plt.imshow(image_np)选中单元格后，点击Run运行
稍等一会，可以看到结果
最后这个识别结果，就是利用我们自己的模型实现的，准确率还是很高的。至此我们实现了Tensorflow object detection API 搭建物体识别模型方法。
