search

cwchenwang / NeRF-SR

NeRF-SR: High-Quality Neural Radiance Fields using Supersampling
https://cwchenwang.github.io/NeRF-SR
138 stars 13 forks source link

question about pretrained model #4

Closed puresky123 closed 2 years ago

puresky123 commented 2 years ago

Hi, I'm very interested in your work ! But when I use command—bash scripts/test_llff_refine.sh to see the effect of the refinement model. It told me that I need a folder named 30_test_vis which your pretrained model is not provided. It seems that I need to retrain the model before refinement. Could you tell me how can I see the result by only using the pretrained model you provide? Thanks a lot !

cwchenwang commented 2 years ago

The 30_test_vis contains the testing poses that we used for creating video. You can directly render it by yourself using the provided test_llff_downX.sh and checkpoints (namely the 30_net_Coarse.pth and 30_net_Fine.pth), which doesn't need to retrain the model. You can also change 30_test_vis to 30_val_vis to see how refinement works on validation poses.

puresky123 commented 2 years ago

Thank you ! I'll close the issue.

puresky123 commented 2 years ago

Hi, I have another question. If I want to test the llff_refinement model, how can I get the i_locs.npz files which were contained in the folder named 30_test_vis. Thank you !

cwchenwang commented 2 years ago

Use the provided warp.py

puresky123 commented 2 years ago

Thank you so much !