Hello, I am attempting to compare performance of several different custom models on the oak-d. All models are in the onnx format and I have been using the online converter to convert them to the required blob format. For some models I get an error on the blob conversion step saying something along the lines of unsupported layer type "HardSigmoid". This is a layer that many of the models I would like to try implement, but it seems it is not supported by the blob conversion. Is there any way to convert these models without changing the underlying network architecture to remove the HardSigmoid layer?
