Zidan Yang

@vincentfung13 same question (especially the Flower code)

> Hi, 请参考这条[回复](https://github.com/LongguangWang/PAM/issues/6#issuecomment-850440294)。 谢谢！

`sudo docker run --gpus all --ipc=host -v {your_path}:/home/CLIPasso --rm -it yaelvinker/clipasso_docker /bin/bash` Just use this

Hi, have you solved this problem?

By the way, what kind of LPIPS metric do you use, vgg or alex?

> Thanks for your interest! We use alex to evaluate LPIPS and compare different methods using the checkpoint of the same epoch. Thank you so much!

the reason seems to be the large value of the parameter "image_size" of the Pytorch3dRasterizer. `rasterizer = Pytorch3dRasterizer(image_size=2**12)` Also, the rasterization process is running on CPU mode, which further slows...

@alvinliu0 Hello, have you updated this issue, I also came across this issue, resulting in one human face in the front view and one human face in the back view.

Cause the value of z is negative in the default camera coordinate.