search

google-research / multinerf

A Code Release for Mip-NeRF 360, Ref-NeRF, and RawNeRF
Apache License 2.0
3.57k stars 338 forks source link

Rendering images from view points not in test set #81

Open Bekci opened 1 year ago

Bekci commented 1 year ago

Hello, the dataset I am using consists of a few images (10-15 per scene). I want to generate a 360 degree video of the scene that consists of images from every angle. It seems like render.py code renders test images only. How can I render images which are not in the test set?

jonbarron commented 1 year ago

eval.py renders test-set views, render.py should render a 360 swivel view.

Bekci commented 1 year ago

But it traverses whole test dataset and renders them to a new image file. Then the video is constructed by the generated images.

I used below command to run:

python -m render \
  --gin_configs=configs/nuscenes.gin \
  --gin_bindings="Config.data_dir = '${DATA_DIR}/${SCENE}'" \
  --gin_bindings="Config.checkpoint_dir = '${CHECKPOINT_DIR}'" \
  --gin_bindings="Config.render_path = True" \
  --gin_bindings="Config.render_path_frames = 480" \
  --gin_bindings="Config.render_dir = '${CHECKPOINT_DIR}/render/'" \
  --gin_bindings="Config.render_video_fps = 60" \
  --logtostderr

Could it be related with my custom dataloader?

ElliotQi commented 1 year ago

Hi, I have the same question. How is it going?