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

How to make bounding box data? #38

Open nicesonnday opened 1 year ago

nicesonnday commented 1 year ago

Hi, I want to train ENeRF with my own data.

My own data constructed on 3*7 fixed grid multi-view cameras.

In this moment, I have to make bounding box data (*.npy) for my own data to train on ENeRF-Outdoor dataset setting.

Please let me know how to make the bounding box (or is there any code for make bounding box?)

haotongl commented 1 year ago

Hi, thanks for your attention!

The 3D bounding box can be defined as very rough hyper-parameter, because our algorithm does not require an accurate 3D bounding box.

If the foreground is mainly human, we recommend using EasyMocap to estimate the 3D key points of the human body, and then determine the bbox based on the 3D key points. https://github.com/zju3dv/EasyMocap

If the foreground and background are relatively clear, you can also use the background matting and visual hull algorithms to determine the bbox. https://github.com/zju3dv/ENeRF/issues/27#issuecomment-1450173304