onnx
alicera opened this issue · comments
Hi SthPhoenix,
How to run the model to onnx?
https://github.com/deepinsight/insightface/tree/master/recognition/partial_fc
thanks
Hi! You need to manually download glint360k_r100FC_1.0 or glint360k_r100FC_0.1 model and symbol files to models/mxnet/glint360k_r100FC... and change rec_model name to the model you selected. Both these models are supported.
-
mxnet model
It show the error
-
pytorch model
Do you have the process the pt model tool?
https://github.com/deepinsight/insightface/tree/master/recognition/arcface_torch
- That's correct, this model isn't part of original insightface python package, used for comparison.
- I can't download it from baidu, so I can't check it )
- I can't download it from baidu, so I can't check it )
You can download it here
That's great! Thanks for the info, I'll check if I can convert it later.
I modify the model name "coordinateReg" in the insight_tester.py
But it show the message no this model
Hi! As it's stated in readme coordinateReg can be converted, but it's inference is not supported.
I run the command "python insight_tester.py" and modify the model name "coordinateReg" in the insight_tester.py.
So the final result will get the onnx model but inference is not suppoerted.
By the way, do you know the coordinateReg project dataset ?
I run the command "python insight_tester.py" and modify the model name "coordinateReg" in the insight_tester.py.
So the final result will get the onnx model but inference is not suppoerted.
No, for conversion you can use /src/converters/build_insight_trt.py
By the way, do you know the coordinateReg project dataset ?
No, I think this question is better to be addressed to authors of the model.
Closing. Added support for original glintr100 model, based on PyTorch model.