Jianzhu Guo
Jianzhu Guo
I did not test it on CentOS. Or, you can disable it to use the original render Sim3DR, it also works.
This repo is not tested for onnxruntime-gpu, so there may be some workloads to implement it but the code change maybe small. Sorry for that I have no time for...
Thanks for your sharing. Hope anyone can provide a detailed comparison.
Converting the module and tensor to gpu device will be ok.
You may contact Xiangyu Zhu to consult the generation details. I guess the landmarks are generated by warping the 2d landmarks of the original image. If the wrap relation between...
You can refer https://github.com/cleardusk/3DDFA_V2/issues/23#issuecomment-716330320
For the NME part, you can refer to the benchmark of 3DDFA, [here](https://github.com/cleardusk/3DDFA/blob/master/benchmark.py). For PIFA, you may refer to the public code in their project page.
You can cut out the eye indices.
It is a misdetection. You can try to modify the confidence threshold [here](https://github.com/cleardusk/3DDFA_V2/blob/master/FaceBoxes/FaceBoxes.py#L18).