google-deepmind / google-deepmind/open_x_embodiment
Observation History Question
Open
- Dominant language
- Jupyter Notebook
- Stars
- 2k
- Forks
- 127
- PR merge metrics
- No merged PRs in 30d
Description
Just a couple of questions about the image observations:
- Should the latest image be the first element of the image sequence? That is, index zero of the (15, 300, 300) image seq?
- What is best practice for the first 15 time steps where there is not an observation history? For instance, at time step 1, should the other 14 image observations be padded with zero arrays or perhaps should the current observation be repeated 14 times?
:)
Contributor guide
Assessment
This issue has not been assessed yet.