optical_flow

Optical Flow Prediction

For optical flow prediction we use the MOVi-F dataset which is explained in challenges/movi. MOVi-F is identical to MOVi-E except that it adds a random amount of motion blur to each video and was rendered in 512x512 resolution (with downscaled variants for 256x256 and 128x128).

Generate single scene with the movi_def_worker.py script:

docker run --rm --interactive \
  --user $(id -u):$(id -g)    \
  --volume "$(pwd):/kubric"   \
  kubricdockerhub/kubruntu    \
  /usr/bin/python3 challenges/movi/movi_def_worker.py \
  --camera=linear_movement
  --max_motion_blur=2.0

See movi_f.py for the TFDS definition / conversion.

Data is located at gs://kubric-public/tfds/movi_f and can be loaded with:

ds = tfds.load("movi_f", data_dir="gs://kubric-public/tfds")