search

zju3dv / ENeRF

SIGGRAPH Asia 2022: Code for "Efficient Neural Radiance Fields for Interactive Free-viewpoint Video"
https://zju3dv.github.io/enerf
Other
413 stars 28 forks source link

Config file for fine-tune zjumocap dataset #12

Closed weiyichang closed 1 year ago

weiyichang commented 1 year ago

Hi,

Thanks for sharing this nice work.

I would like know that how to fine-tune model for zjumocap dataset.

I have modified the config from zjumocap_eval.yaml, but the results are worse than the pre-trained model.

Do you have suggestions ?

Thanks !!!

haotongl commented 1 year ago

I'll check it out, please be patient.

haotongl commented 1 year ago

I have updated the corresponding config file and README. If everything is set up, the following results can be obtained by tensorboard --logdir=$workspace/record --bind_all: image

weiyichang commented 1 year ago

Thanks a lot, I will try it. One more thing I would like to confirm is about the epoch number in train part, Is it necessary to train with 100 epoches ? (I will try it by myself, but I also would like to know your experience) Thanks !!!

haotongl commented 1 year ago

From my experience, there will be no obvious visual improvement after several (5) epochs. But the metric may fully converge after 40 epochs.

weiyichang commented 1 year ago

Thanks for your quick response. I will try it :)