search

postech-ami / CLIP-Actor

[ECCV'22] Official PyTorch Implementation of "CLIP-Actor: Text-Driven Recommendation and Stylization for Animating Human Meshes"
https://clip-actor.github.io
MIT License
108 stars 5 forks source link

about the settings in paper #3

Closed lai-pf closed 2 years ago

lai-pf commented 2 years ago

I saw the text prompt "Tony Stark wearing blue suit walking forwards" in this condition, the result of Text2mesh should like this image but I got this , with the default settings of ninja demo in text2mesh image can you tell me about the setting of the experiment between Text2mesh and Clip-Actor while using this text promt.

Youwang-Kim commented 2 years ago

@lai-pf Hi there, we used the default configuration of the Text2Mesh when we get the results. Please note that there are some randomness in Text2Mesh and CLIP-Actor, mainly during the optimization process. Since we did not fix the cuda seed (for generating the samples), the gradient computed from your experiment and our paper's result may differ each other.

Youwang-Kim commented 2 years ago

Please reopen this issue if you need more help regarding this.