search

Le-Xiaohuai-speech / DPCRN_DNS3

Implementation of paper "DPCRN: Dual-Path Convolution Recurrent Network for Single Channel Speech Enhancement"
180 stars 40 forks source link

Convert to ONNX format #28

Open kevin550029 opened 1 year ago

kevin550029 commented 1 year ago

Hi, thank you for sharing this great work and code.

I noticed that there is a code about dpcrn_onnx_demo in the real_time_inference folder.

Have you ever transferred DPCRN model into ONNX format?

I am trying to remove the stft/ifft layer and convert real-time model into ONNX using keras2onnx.

but I got this warning, and the generated onnx seems wrong

WARN: No corresponding ONNX op matches the tf.op node decoder_dconv_5/conv2d_transpose of type Conv2DBackpropInput The generated ONNX model needs run with the custom op supports.

Can you give some advice for onnx conversion steps? Thank you.

SEMLLYCAT commented 1 year ago

Thank you for your excellent work and sharing. When I tried to replicate your work, I found that dpcrn.onnx or model.onnx was missing in run Run. cpp. I don't know whether this onnx was converted from dpcrn_stateful_model.tflite. I tried tflite2onnx. However, various errors will occur constantly. Could you please provide the onnx or how to convert the method to get onnx? We are looking forward to your reply very much.

zeynepgulhanuslu commented 10 months ago

Hi,

Do you manage to convert model to onnx?

Thanks

ronalbo commented 4 days ago

Dear author, hello. I hope you can solve the problem with onnx. I would be extremely grateful