I have referred to this link below which suggests trimming the batchedNMS branch and using the NMS plugin later. Is this the best practice?
Can we change the yolo_export code in the backend without NMS branch while converting yolov3/v4 model to onnx, how do we do it on colab?
Yes, you can also change the export code.
It is not very convenient in colab. But you can still refer to Unable to export hdf5 to etlt after Tao Training on Colab - #9 by Morganh and Unable to export hdf5 to etlt after Tao Training on Colab - #11 by Morganh about how to login docker and modify code.
How do we edit yolov4 export.py code in colab to convert the hdf5 model to onnx without batchedNMS branch instead of trimming the model?
As mentioned above, you can change code via sed, see Unable to export hdf5 to etlt after Tao Training on Colab - #11 by Morganh.
But it’s not convenient. Suggest you to modify code locally, then copy into google drive.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.