search

VITA-Group / FSGS

[ECCV 2024]"FSGS: Real-Time Few-Shot View Synthesis using Gaussian Splatting", Zehao Zhu*, Zhiwen Fan*, Yifan Jiang, Zhangyang Wang
Other
417 stars 34 forks source link

Question about generating random poses #36

Open ghost opened 9 months ago

ghost commented 9 months ago

Hi, thanks for the fancy work. I have tried running the code and it works well. However, I am confused that It seems that the generation of pseudo poses is the average of all training view poses with noise, but the description on Page 4 of this paper is "The synthesized view is sampled from the two closest training views in Euclidean space, calculating the averaged camera orientation and interpolating a virtual one between them." I can't find the corresponding code related to generating random poses described in the paper. Please confirm if my understanding is correct, or if there is any misunderstanding of the code.

trainable commented 6 months ago

@kqjt0e11gy0jkgmy Hi, could you share your configuration information? For example, Python version, PyTorch version, CUDA version, etc. I haven't resolved the memory error issue despite trying the solutions mentioned above.

ghost commented 5 months ago

@kqjt0e11gy0jkgmy Hi, could you share your configuration information? For example, Python version, PyTorch version, CUDA version, etc. I haven't resolved the memory error issue despite trying the solutions mentioned above.

I run the FSGS code on the RTX3090 with python3.8, torch1.12, CUDA11.3 (I am not sure about it since recently I did't run it) However, I don't think that the memory error issue is caused by the environment configuration?

sailor-z commented 4 months ago

Hi, thanks for the fancy work. I have tried running the code and it works well. However, I am confused that It seems that the generation of pseudo poses is the average of all training view poses with noise, but the description on Page 4 of this paper is "The synthesized view is sampled from the two closest training views in Euclidean space, calculating the averaged camera orientation and interpolating a virtual one between them." I can't find the corresponding code related to generating random poses described in the paper. Please confirm if my understanding is correct, or if there is any misunderstanding of the code.

I'm also curious about this. I noticed that the follow-up paper uses the same introduction about the pseudo-view generations. Could you please look into this since it is quite confusing?