facebookresearch / facebookresearch/sam2

Redundant arguments in Hiera ?

Open
#290 10 comments 0 reactions 0 assignees View on GitHub
Dominant language
Jupyter Notebook
Stars
19.9k
Forks
2.5k
PR merge metrics
No merged PRs in 30d

Description

I'm a little confused by the default arguments passed to the Hiera class. https://github.com/facebookresearch/segment-anything-2/blob/7e1596c0b6462eb1d1ba7e1492430fed95023598/sam2/modeling/backbones/hieradet.py#L184

It seems like `window_spec` is hiding two things at once.
In the first window attention blocks, They have respectively 8 and 4 tokens (That's window attention).
After those, they will have 14 and 7 tokens (That's all the tokens left, so that is global attention).

But then it looks redundant with the `global_att_blocks` parameters.

Can you help me understand those 3 parameters, `window_spec`, `global_att_blocks` and `window_pos_embed_bkg_spatial_size`.

Thanks!

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.