search

Fictionarry / ER-NeRF

[ICCV'23] Efficient Region-Aware Neural Radiance Fields for High-Fidelity Talking Portrait Synthesis
https://fictionarry.github.io/ER-NeRF/
MIT License
985 stars 131 forks source link

inference fps are 17fps in RTX3090 #40

Open xfguo-ucas opened 12 months ago

xfguo-ucas commented 12 months ago

hi, I test the model(video: obama.mp4) in my rtx3090, the inference fps only 17fps, but paper shows fps are 34 in rtx3080ti. Can you give me some advice?

Fictionarry commented 12 months ago

Little information is given in your words, but first I think you should focus on the FPS during runtime (rather than the latest number) and check whether it can achieve the reported speed. If it can't, the problem may be in the environment or other hardware.

xfguo-ucas commented 12 months ago

Little information is given in your words, but first I think you should focus on the FPS during runtime (rather than the latest number) and check whether it can achieve the reported speed. If it can't, the problem may be in the environment or other hardware.

ok, thank for your reply.

kike-0304 commented 10 months ago

Little information is given in your words, but first I think you should focus on the FPS during runtime (rather than the latest number) and check whether it can achieve the reported speed. If it can't, the problem may be in the environment or other hardware.

ok, thank for your reply.

May I ask what is the inference time for each frame of your image? The inference time for each frame of my A800 graphics card is 42 ms.