[神经网络]一步一步使用Mobile-Net完成视觉识别（五）

1.环境配置

2.数据集获取

3.训练集获取

4.训练

5.调用测试训练结果

6.代码讲解

　　本文是第五篇，讲解如何调用测试训练结果。

上一篇中我们输出了训练的模型，这一篇中我们通过调用训练好的模型来完成测试工作。

在object_detection目录下创建test.py并输入以下内容：

import os
import cv2
import numpy as np
import tensorflow as tf
import sys
sys.path.append("..")
from utils import label_map_util
from utils import visualization_utils as vis_util

ENERMY = 2 # 1 代表蓝色方，2 代表红色方 ，设置蓝色方为敌人
DEBUG = False
THRE_VAL = 0.2

PATH_TO_CKPT ='/home/xueaoru/models/research/inference_graph_v2/frozen_inference_graph.pb'
PATH_TO_LABELS = '/home/xueaoru/models/research/object_detection/car_label_map.pbtxt'
NUM_CLASSES = 2
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)
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='')

    sess = tf.Session(graph=detection_graph)
image_tensor = detection_graph.get_tensor_by_name('image_tensor:0')
detection_boxes = detection_graph.get_tensor_by_name('detection_boxes:0')
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')

def video_test():
    #cap = cv2.VideoCapture(1)
    cap = cv2.VideoCapture("/home/xueaoru/下载/RoboMaster2.mp4")
    while(1):
        time = cv2.getTickCount()
        ret, image = cap.read()
        if ret!= True:
            break
        image_expanded = np.expand_dims(image, axis=0)#[1,w,h,3]

        (boxes, scores, classes, num) = sess.run(
        [detection_boxes, detection_scores, detection_classes, num_detections],
        feed_dict={image_tensor: image_expanded})
        #print(np.squeeze(classes).astype(np.int32))
        #print(np.squeeze(scores))
        #print(np.squeeze(boxes))
        vis_util.visualize_boxes_and_labels_on_image_array(
        image,
        np.squeeze(boxes),
        np.squeeze(classes).astype(np.int32),
        np.squeeze(scores),
        category_index,
        use_normalized_coordinates=True,
        line_thickness=8,
        min_score_thresh=0.4)

        cv2.imshow('Object detector', image)
        key = cv2.waitKey(1)&0xff
        time = cv2.getTickCount() - time
        print("处理时间："+str(time*1000/cv2.getTickFrequency()))
        if key ==27:
            break
    cv2.destroyAllWindows()
def pic_test():
    image = cv2.imread("/home/xueaoru/models/research/images/image12.jpg")
    image_expanded = np.expand_dims(image, axis=0)  # [1,w,h,3]

    (boxes, scores, classes, num) = sess.run(
        [detection_boxes, detection_scores, detection_classes, num_detections],
        feed_dict={image_tensor: image_expanded})

    if DEBUG:
        vis_util.visualize_boxes_and_labels_on_image_array(
        image,
        np.squeeze(boxes),
        np.squeeze(classes).astype(np.int32),
        np.squeeze(scores),
        category_index,
        use_normalized_coordinates=True,
        line_thickness=8,
        min_score_thresh=0.80)
    else:
        score = np.squeeze(scores)
        max_index = np.argmax(score)
        score = score[max_index]
        detected_class = np.squeeze(classes).astype(np.int32)[max_index]
        if score > THRE_VAL and detected_class == ENERMY:
            box = np.squeeze(boxes)[max_index]#(ymin,xmin,ymax,xmax)
            h,w,_ = image.shape
            min_point = (int(box[1]*w),int(box[0]*h))
            max_point = (int(box[3]*w),int(box[2]*h))
            cv2.rectangle(image,min_point,max_point,(0,255,255),2)

    cv2.imshow('Object detector', image)
    cv2.waitKey(0)
    cv2.destroyAllWindows()
video_test()

好了，暂时就先这样吧，最后一篇详细讲解包括通过这些识别到的框到最后计算炮台偏转角度的代码。这段代码的讲解也放在后面。