Error in using tensorflow model with opencv dnn module. #17364
Description
Hi.
Im working on a road segmentation project with tensorflow. I searched and find this usefull repo :
https://github.com/MaybeShewill-CV/lanenet-lane-detection
Author has provided a link to pre-trained checkpoints file as this :
https://www.dropbox.com/sh/tnsf0lw6psszvy4/AAA81r53jpUI3wLsRW6TiPCya?dl=0
I downloaded this check points and froze them with freeze_lanenet_model.py script in the mnn_project folder of project and now i have a model with this name : lanenet.pb
freeze_lanenet_model main function is this :
def convert_ckpt_into_pb_file(ckpt_file_path, pb_file_path):
:param ckpt_file_path:
:param pb_file_path:
:return:
"""
# construct compute graph
with tf.variable_scope('lanenet'):
input_tensor = tf.placeholder(dtype=tf.float32, shape=[1, 256, 512, 3], name='input_tensor')
net = lanenet.LaneNet(phase='test', net_flag='vgg')
binary_seg_ret, instance_seg_ret = net.inference(input_tensor=input_tensor, name='lanenet_model')
with tf.variable_scope('lanenet/'):
binary_seg_ret = tf.cast(binary_seg_ret, dtype=tf.float32)
binary_seg_ret = tf.squeeze(binary_seg_ret, axis=0, name='final_binary_output')
instance_seg_ret = tf.squeeze(instance_seg_ret, axis=0, name='final_pixel_embedding_output')
# create a session
saver = tf.train.Saver()
sess_config = tf.ConfigProto()
sess_config.gpu_options.per_process_gpu_memory_fraction = 0.85
sess_config.gpu_options.allow_growth = False
sess_config.gpu_options.allocator_type = 'BFC'
sess = tf.Session(config=sess_config)
with sess.as_default():
saver.restore(sess, ckpt_file_path)
converted_graph_def = tf.graph_util.convert_variables_to_constants(
sess,
input_graph_def=sess.graph.as_graph_def(),
output_node_names=[
'lanenet/input_tensor',
'lanenet/final_binary_output',
'lanenet/final_pixel_embedding_output'
]
)
with tf.gfile.GFile(pb_file_path, "wb") as f:
f.write(converted_graph_def.SerializeToString())`
Now i wanna use this saved model in opencv dnn module like this :
Net net = readNetFromTensorflow("lanenet.pb");
But when i run the code , i have this error :
OpenCV(4.2.0) /home/eagle-soft/opencv/modules/dnn/src/tensorflow/tf_importer.cpp:544: error: (-2:Unspecified error) Input layer not found: lanenet_model/vgg_frontend/vgg16_decode_module/instance_seg_decode/decode_stage_1_fuse/deconv/deconv/Shape in function 'connect'
I searched a lot but couldn't find a same issue and solution.
Please give me some workaround to solve this annoying problem.
Thanks in advande.