Export Faster R-CNN to ONNX

Hi, I am trying to convert my trained faster R-CNN model to ONNX in order to export to other framework. However, i get the error as below:

Error using

nnet.internal.cnn.onnx.exportONNXNetwork>iValidateNetwork (line

35)

Network argument must be a DAGNetwork or a SeriesNetwork.

Error in nnet.internal.cnn.onnx.exportONNXNetwork>iValidateInputs

(line 28)

NNTNetwork = iValidateNetwork(NNTNetwork);

Error in nnet.internal.cnn.onnx.exportONNXNetwork (line 7)

[NNTNetwork, Filename, NetworkName] = iValidateInputs(NNTNetwork,

Filename, varargin{:});

Error in exportONNXNetwork (line 34)

nnet.internal.cnn.onnx.exportONNXNetwork(Network, filename,

varargin{:});

Is it mean the faster R-CNN model not a DAGnetwork? Is there any way to solve this problem? Appreciate for the help

[Jon N]

I too have been trying to export a faster R-CNN model to ONNX. Just a basic model that was trained with trainFasterRCNNObjectDetector. MATLAB throws errors that some layers aren't supported, and thus I cannot use the exported ONNX model. At this point I am wondering what is the point of using MATLAB for deep learning if I cannot deploy these models in other frameworks.

[Ting Su]

Hi, Hui and Jon,

We are looking at supporting the export of objection detection models.