search

skhu101 / SHERF

Code for our ICCV'2023 paper "SHERF: Generalizable Human NeRF from a Single Image"
Other
297 stars 10 forks source link

Multiple frames per subject during training? #2

Closed pansanity666 closed 1 year ago

pansanity666 commented 1 year ago

Hi, It seems that multiple frames for each subject is required during training since the target view is different from the input view in Fig. 2. Did you try one frame per subject during training?

skhu101 commented 1 year ago

Hi, we currently use multiple frames (different poses) for each subject during training as we hope to recover the 3D human in the canonical space. It is interesting to try only using one frame (pose) per subject during training.

pansanity666 commented 1 year ago

Did you use random cameras together with random poses when sampling multiple frames for a subject?

skhu101 commented 1 year ago

Hi, we randomly sample one target and one input camera view from the same pose of the same subject to construct the pair data during training.