Open kyanagida3407 opened 1 year ago
Hi for forward-facing style scenes like LLFF for example, you would need to change the "render_poses" https://github.com/googleinterns/IBRNet/blob/master/ibrnet/data_loaders/llff_test.py#L53. The render_poses will then be used in the eval.py file to render the image.
Thank you very much. I confirmed that I can render with free paths by rewriting the part you mentioned.
I have one more question, and I would be glad if you could answer it for me.
When I use my own LLFF image as input image, the rendering result is different from what I expect. I put the attached file in data/nerf_llff_data/ and changed config/eval_llff.txt to render it, and the other attached movie was rendered. The trained model is the distributed pretrained/model_25500.pth.
If there is anything special I need to do when rendering my own data, please let me know.
https://drive.google.com/drive/folders/14dbeYxBF6COqBBW5fKlGK-l_ynV1Yn-h?usp=sharing https://drive.google.com/file/d/1H0CViqXbNWIewWALVUE_-bMW4CUNf64W/view?usp=sharing
Hi. I would like to know which parameters can be changed to render the video with an arbitrary render pass. Thank you in advance.