search

chungyiweng / humannerf

HumanNeRF turns a monocular video of moving people into a 360 free-viewpoint video.
MIT License
786 stars 86 forks source link

to confirm the experiment setting of ZJUmocap #71

Open yz-cnsdqz opened 1 year ago

yz-cnsdqz commented 1 year ago

Dear authors,

Thanks for sharing this nice work. After reading your code and paper, I would like to confirm on the training frames in the ZJUmocap experiment.

In your configure files, the training_view means the camera ID rather than camera name. The max_frames means the frames starting from the first frame (frame index 0, and max_frames=-1 means using all frames, eg in https://github.com/chungyiweng/humannerf/blob/d1a623c0e43dc15f112ed0b62d4a8fa487c7e5d2/tools/prepare_zju_mocap/390.yaml#L7 For testing, you used the same frame IDs from all other views.

Am I right? Please let me know if I am wrong. Great thanks!

best regards, and looking forward to your reply.

Dipankar1997161 commented 1 year ago

frame_ids dont change irrespective of train or test

Keep in mind, Humannerf, uses a single camera from zju-mocap, which is mostly camera 1.

yz-cnsdqz commented 1 year ago

Thanks! Does max_frames: -1 mean using all frames in train or teat?

Dipankar1997161 commented 1 year ago

Thanks! Does max_frames: -1 mean using all frames in train or teat?

In both cases. When you run the train.py file, you will see the total number of frames will be equal to that of the actual frames in mocap