We have a model using conv, deconv, concat layers. But the output from ncsdkv2 and v1 are both incorrect.
And if we are not using explicit_concat parameter when convert the model, the output will become NaN also.
But even with explicit_concat parameters enable, the result is still completely wrong.
We are using caffe model and have a python caffe test application for the expected output.
How we can submit the model for further diagnostic ?
