我已按照this教程使用Tensorflow GPU重新训练MobileNet SSD V1，并在使用GPU训练后获得0.5损失（下面有关配置的更多信息）并获得model.ckpt。
这是我用于训练的命令：
python ../models/research/object_detection/legacy/train.py --logtostderr --train_dir=./data/ --pipeline_config_path=./ssd_mobilenet_v1_pets.config
这是冻结命令（生成pb文件）：
python ../models/research/object_detection/export_inference_graph.py --input_type image_tensor --pipeline_config_path ./ssd_mobilenet_v1_pets.config --trained_checkpoint_prefix ./data/model.ckpt-1407 --output_directory ./data/
这是我使用冷冻pb和pbtxt时得到的错误：
Traceback (most recent call last):
File "Object_detection_image.py", line 29, in <module>
cvOut = cvNet.forward()
cv2.error: OpenCV(3.4.3) C:\projects\opencv-python\opencv\modules\dnn\src\dnn.cpp:565: error: (-215:Assertion failed) inputs.size() == requiredOutputs in function 'cv::dnn::experimental_dnn_34_v7::DataLayer::getMemoryShapes'
这是我使用的Object_detection_image.py文件：
import cv2 as cv
import os
import time
import logging
logger = logging.getLogger()
fh = logging.FileHandler('xyz.log')
fh.setLevel(logging.DEBUG)
logger.addHandler(fh)
cvNet = cv.dnn.readNetFromTensorflow('frozen_inference_graph.pb', 'object_detection.pbtxt')
dir_x = "C:\\Users\\Omen\\Desktop\\LP_dataset\\anno"
for filename in os.listdir(dir_x):
print(filename)
if not (filename.endswith(".png") or filename.endswith(".jpg")):
continue
print('daz')
img = cv.imread(os.path.join(dir_x,filename))
img = cv.resize(img, (300,300))
#cv.imshow('i',img)
#cv.waitKey(0)
img = cv.cvtColor(img,cv.COLOR_BGR2GRAY)
img = cv.cvtColor(img,cv.COLOR_GRAY2RGB)
rows = img.shape[0]
cols = img.shape[1]
#cvNet.setInput(cv.dnn.blobFromImage(img, size=(cols,rows), swapRB=True, crop=False))
cvNet.setInput(cv.dnn.blobFromImage(img, size=(300, 300), crop=False))
t0 = time.time()
cvOut = cvNet.forward()
print(time.time() - t0)
for detection in cvOut[0,0,:,:]:
score = float(detection[2])
#print(score)
if score > 0.80:
left = detection[3] * cols
top = detection[4] * rows
right = detection[5] * cols
bottom = detection[6] * rows
cv.rectangle(img, (int(left), int(top)), (int(right), int(bottom)), (23, 230, 210), thickness=2)
cv.imshow('img', img)
cv.waitKey(0)
这是pbtxt文件（我也试过导出的pbtxt并从pb生成pbtxt但不工作）：
item {
id: 1
name: 'licenseplate'
}
配置：
您正在使用的模型的顶级目录是什么：object_detetion
我写过自定义代码：没有
操作系统平台和分发：win10
TensorFlow安装自：二进制
TensorFlow GPU版本：1.13.0
CUDA / cuDNN版本：10
GPU型号：1050 GTX
我可以提供你问的任何文件，请帮帮我。在tensorflow的github中，他们告诉我在Stackoverflow中询问...
更新：
由于答案，我解决了问题，这里是cvOut的内容：
[[[[-0.00476191 -0.00361736 0. 0.25361738 -0.07576995
0.03405379 0.40910327]
[ 0.21594621 0.04544836 0. 0.28788495 0.30689242
-0.13025634 0.05074273]
[ 0.46358964 0.19925728 0. -0.09778295 0.26563603
0.34778297 -0.02014329]
[-0.01515752 0.3534766 0. 0.32857144 -0.00361736
0.67142856 0.25361738]
[ 0.25756338 0.03405379 0. 0.21594621 0.3787817
-0.05689242 0.6212183 ]
[ 0.30689242 0.203077 0. 0.796923 0.19925728
0.40103063 -0.09778295]
[ 0.5989694 0.34778297 0. -0.01515752 0.68680996
0.26515752 0.66190475]
[-0.00361736 1.0047619 0. 0.59089667 0.03405379
1.0757699 0.21594621]
[ 0.712115 -0.05689242 0. 0.30689242 0.53641033
0.05074273 1.1302563 ]
[ 0.19925728 0.7343639 0. 0.93230265 0.34778297
0.64652336 -0.01515752]
[ 1.0201433 0.26515752 0. 0.24638264 0.33809522
0.50361735 -0.07576995]
[ 0.2840538 0.40910327 0. 0.04544836 0.19310758
0.28788495 0.5568924 ]
[-0.13025634 0.30074272 0. 0.44925728 0.06769729
0.15221705 0.26563603]
[ 0.59778297 -0.02014329 0. 0.3534766 0.5151575
0.32857144 0.24638264]
[ 0.67142856 0.50361735 0. 0.2840538 0.7424366
0.4659462 0.3787817 ]
[ 0.19310758 0.6212183 0. 0.203077 0.30074272
0.796923 0.44925728]
[ 0.40103063 0.15221705 0. 0.59778297 0.31319004
0.23484248 0.68680996]
[ 0.5151575 0.66190475 0. 1.0047619 0.50361735
0.59089667 0.2840538 ]
[ 1.0757699 0.4659462 0. 0.19310758 0.95455164
0.5568924 0.53641033]
[ 0.30074272 1.1302563 0. 0.7343639 0.15221705
0.93230265 0.59778297]
[ 0.64652336 0.23484248 0. 0.5151575 -0.00476191
0.49638262 0.33809522]
[ 0.75361735 -0.07576995 0. 0.40910327 0.7159462
0.04544836 0.44310758]
[ 0.28788495 0.8068924 0. 0.55074275 0.46358964
0.69925725 0.06769729]
[ 0.40221703 0.26563603 0. -0.02014329 0.48484248
0.3534766 0.7651575 ]
[ 0.32857144 0.49638262 0. 0.75361735 0.25756338
0.5340538 0.7424366 ]
[ 0.7159462 0.3787817 0. 0.6212183 0.8068924
0.203077 0.55074275]
[ 0.796923 0.69925725 0. 0.40221703 0.5989694
0.84778297 0.31319004]
[ 0.48484248 0.68680996 0. 0.66190475 0.49638262
1.0047619 0.75361735]
[ 0.59089667 0.5340538 0. 0.7159462 0.712115
0.44310758 0.95455164]
[ 0.8068924 0.53641033 0. 1.1302563 0.69925725
0.7343639 0.40221703]
[ 0.93230265 0.84778297 0. 0.48484248 1.0201433
0.7651575 -0.00476191]
[ 0.74638265 0.33809522 0. -0.07576995 0.7840538
0.40910327 0.9659462 ]
[ 0.04544836 0.6931076 0. 1.0568924 -0.13025634
0.80074275 0.46358964]
[ 0.94925725 0.06769729 0. 0.26563603 1.0977829
-0.02014329 0.7348425 ]
[ 0.3534766 1.0151576 0. 0.74638265 0.67142856
1.0036174 0.25756338]
[ 0.7840538 0.7424366 0. 0.3787817 0.6931076
0.6212183 1.0568924 ]
[ 0.203077 0.80074275 0. 0.94925725 0.40103063
0.65221703 0.5989694 ]
[ 1.0977829 0.31319004 0. 0.68680996 1.0151576
0.66190475 0.74638265]
[ 1.0047619 1.0036174 0. 0.7840538 1.0757699
0.9659462 0.712115 ]
[ 0.6931076 0.95455164 0. 0.53641033 0.80074275
1.1302563 0.94925725]
[ 0.7343639 0.65221703 0. 1.0977829 0.64652336
0.7348425 1.0201433 ]
[ 1.0151576 0.1 0. 0.2 0.2
0.1 0.1 ]
[ 0.2 0.2 0. 0.1 0.2
0.2 0.1 ]
[ 0.1 0.2 0. 0.1 0.1
0.2 0.2 ]
[ 0.1 0.1 0. 0.2 0.1
0.1 0.2 ]
[ 0.2 0.1 0. 0.2 0.2
0.1 0.1 ]
[ 0.2 0.2 0. 0.1 0.2
0.2 0.1 ]
[ 0.1 0.2 0. 0.1 0.1
0.2 0.2 ]
[ 0.1 0.1 0. 0.2 0.1
0.1 0.2 ]
[ 0.2 0.1 0. 0.2 0.2
0.1 0.1 ]
[ 0.2 0.2 0. 0.1 0.2
0.2 0.1 ]
[ 0.1 0.2 0. 0.1 0.1
0.2 0.2 ]
[ 0.1 0.1 0. 0.2 0.1
0.1 0.2 ]
[ 0.2 0.1 0. 0.2 0.2
0.1 0.1 ]
[ 0.2 0.2 0. 0.1 0.2
0.2 0.1 ]
[ 0.1 0.2 0. 0.1 0.1
0.2 0.2 ]
[ 0.1 0.1 0. 0.2 0.1
0.1 0.2 ]
[ 0.2 0.1 0. 0.2 0.2
0.1 0.1 ]
[ 0.2 0.2 0. 0.1 0.2
0.2 0.1 ]
[ 0.1 0.2 0. 0.1 0.1
0.2 0.2 ]
[ 0.1 0.1 0. 0.2 0.1
0.1 0.2 ]
[ 0.2 0.1 0. 0.2 0.2
0.1 0.1 ]
[ 0.2 0.2 0. 0.1 0.2
0.2 0.1 ]
[ 0.1 0.2 0. 0.1 0.1
0.2 0.2 ]
[ 0.1 0.1 0. 0.2 0.1
0.1 0.2 ]
[ 0.2 0.1 0. 0.2 0.2
0.1 0.1 ]
[ 0.2 0.2 0. 0.1 0.2
0.2 0.1 ]
[ 0.1 0.2 0. 0.1 0.1
0.2 0.2 ]
[ 0.1 0.1 0. 0.2 0.1
0.1 0.2 ]
[ 0.2 0.1 0. 0.2 0.2
0.1 0.1 ]
[ 0.2 0.2 0. 0.1 0.2
0.2 0.1 ]
[ 0.1 0.2 0. 0.1 0.1
0.2 0.2 ]
[ 0.1 0.1 0. 0.2 0.1
0.1 0.2 ]
[ 0.2 0.1 0. 0.2 0.2
0.1 0.1 ]
[ 0.2 0.2 0. 0.1 0.2
0.2 0.1 ]
[ 0.1 0.2 0. 0.1 0.1
0.2 0.2 ]
[ 0.1 0.1 0. 0.2 0.1
0.1 0.2 ]
[ 0.2 0.1 0. 0.2 0.2
0.1 0.1 ]
[ 0.2 0.2 0. 0.1 0.2
0.2 0.1 ]
[ 0.1 0.2 0. 0.1 0.1
0.2 0.2 ]
[ 0.1 0.1 0. 0.2 0.1
0.1 0.2 ]
[ 0.2 0.1 0. 0.2 0.2
0.1 0.1 ]
[ 0.2 0.2 0. 0.8479438 0.67317617
0.5581815 0.1778345 ]
[-0.9215721 1.5896183 0. 0.6099795 0.5955366
-0.46569395 -0.8461083 ]
[ 1.6129647 1.4244858 0. 0.5209342 0.17585325
-0.8687666 1.7872683 ]
[ 1.3389692 0.8533131 0. -0.00590521 -0.7195761
1.6236191 1.1828533 ]
[ 1.1838211 0.6728102 0. -0.785988 1.2751837
1.1616383 0.933811 ]
[ 0.4684658 0.2719049 0. 1.2093123 0.66612804
0.66964823 0.55971766]
[ 0.17104894 -1.0688283 0. 0.6494252 0.6844874
0.66586125 0.01329695]
[-1.2607187 -0.22749203 0. -0.8741171 -0.9443728
-0.9659323 -0.03422031]
[-0.0364061 0.54829746 0. 0.6263525 0.66758543
0.04167109 -0.11780822]
[ 0.48400337 0.4685324 0. -0.04594427 0.02469592
-0.3487326 0.08831279]
[ 0.4161314 0.23332608 0. -0.13553022 -0.31008872
0.04969648 0.5674252 ]
[ 0.36492363 -0.07475745 0. -0.03859219 0.2016789
-0.39845943 -0.07058203]
[-0.08173721 0.1720942 0. 0.02323131 0.07122216
0.07469177 0.12792486]
[-0.24689877 0.196296 0. 0.5564647 0.535513
0.22528338 -0.37152448]
[-1.7235181 -1.8204601 0. -1.5040898 -1.8099409
-1.8550183 -1.1855855 ]
[-1.6341007 -1.3448519 0. -1.6656716 -1.6564709
-1.2735447 -1.3357594 ]
[-1.2829769 -1.2869868 0. -1.6657944 -1.4066424
-1.4230443 -1.4196167 ]
[-1.3691044 -1.656098 0. -1.4339573 -1.5685135
-1.633306 -1.4437945 ]]]]
错误是由错误的输入.pbtxt文件传递到函数readNetFromTensorflow引起的，因为.pbtxt必须由tf_text_graph_ssd.py生成，如描述here：
Run this script to get a text graph of SSD model from TensorFlow Object Detection API. Then pass it with .pb file to cv::dnn::readNetFromTensorflow function.
对于其他模型，如faster r-cnn和mask r-cnn，也有相应的脚本。
PS：我刚发现有一个非常好的官方教程here.