search

google-research / kubric

A data generation pipeline for creating semi-realistic synthetic multi-object videos with rich annotations such as instance segmentation masks, depth maps, and optical flow.
Apache License 2.0
2.32k stars 228 forks source link

How to generate the whole dataset #243

Open Wangt-CN opened 2 years ago

Wangt-CN commented 2 years ago

Thanks for this awesome system. I have read the docs and readme. It seemed that the provided scripts of movi is for generating one scene:

docker run --rm --interactive \ --user $(id -u):$(id -g) \ --volume "$(pwd):/kubric" \ kubricdockerhub/kubruntu \ /usr/bin/python3 challenges/movi/movi_ab_worker.py \ --objects_set=clevr \ --background=clevr \ --camera=clevr

I wonder how to generate the whole dataset? Need I write a new script to call such single scene generation command again and again?

Wangt-CN commented 2 years ago

Besides, I understand that movi_def_worker.py is used to generate a single scene. But what is the role of the movi_d.py? Is it used to generate the whole dataset? Thanks a lot for clarifying my doubts.

MrXandbadas commented 2 years ago

Hey there! movi_d.py to my understanding is just another worker file

There are lots of ways you can go about generating a whole dataset. You could write a new script to call a new scene generation command again and again. OR you could attempt to build a worker file that creates all the data you need.