During import of a ConcatV2 TensorFlow layer, constant layers are ignored #12931
Description
System information (version)
- OpenCV => 4.0.0 - pre ❔
- Operating System / Platform => Ubuntu 16.04 ❔
- Compiler => GCC ❔
Detailed description
During import of a ConcatV2 TensorFlow layer, input constant layers are ignored
For example:
lambda_4/Variable is constant layer, but it is not found in the graph
node {
name: "lambda_4/concatenate_2/concat"
op: "ConcatV2"
input: "max_pooling2d_2/MaxPool"
input: "lambda_4/Variable"
input: "lambda_4/concatenate_2/concat/axis"
attr {
key: "N"
value {
i: 2
}
}
attr {
key: "T"
value {
type: DT_FLOAT
}
}
attr {
key: "Tidx"
value {
type: DT_INT32
}
}
}
Steps to reproduce
We can see that code (populateNet function in the tf_importer.cpp):
for (int ii = from; ii < to; ii++)
{
Pin inp = parsePin(layer.input(ii));
if (layer_id.find(inp.name) == layer_id.end())
CV_Error(Error::StsError, "Input layer not found: " + inp.name);
connect(layer_id, dstNet, inp, id, ii - from);
}
We are looking for it in the layer_id only, but all const layers place in the value_id.