google-deepmind / google-deepmind/deepmind-research
Perceiver IO: Optical Flow TRAIN_SIZE prediction
Open
- Dominant language
- Jupyter Notebook
- Stars
- 15.2k
- Forks
- 2.9k
- PR merge metrics
- No merged PRs in 30d
Description
In the optical_flow colab, if the size of the images equals the TRAIN_SIZE, the same image patch is predicted 4 times because the compute_grid_indices produces 4 times (0,0).
Contributor guide
Assessment
This issue has not been assessed yet.