Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

How to reduce voxel size #6

Open
TianhaoFu opened this issue May 5, 2022 · 4 comments
Open

How to reduce voxel size #6

TianhaoFu opened this issue May 5, 2022 · 4 comments

Comments

@TianhaoFu
Copy link

hi,

nice repo!

i wonder to know how to reduce the voxel size?

thanks!

@leeminyoung96
Copy link

Hello @TianhaoFu,

Could this be what you are looking for?

Have a look at the voxel_size in the config. (configs/vista/vista.py)

voxel_VISTA

https://github.com/Gorilla-Lab-SCUT/VISTA/blob/main/configs/vista/vista.py

@TianhaoFu
Copy link
Author

hi, when i reduce the voxel size in this config

i found the ret_bev feature map and ret_rv feature from the backbone output have different channel, which mean i could't set the neck in_features channels

can you helps me?

thanks!

Hello @TianhaoFu,

Could this be what you are looking for?

Have a look at the voxel_size in the config. (configs/vista/vista.py)

voxel_VISTA

https://github.com/Gorilla-Lab-SCUT/VISTA/blob/main/configs/vista/vista.py

@leeminyoung96

@leeminyoung96
Copy link

@AndlollipopDE, would you be able to help @TianhaoFu ?
Thank you

@AndlollipopDE
Copy link
Member

To reduce the voxel size, you can simply modify the configuration, but it is necessary to modify the configuration (input_channel, bev/rv_size, etc) of the attention module in ohs.py because the change in voxel size will lead to the change in the shape of 3D feature map produced by the 3D backbone, which will further influence the shape and the channel number of collapsed BEV and RV feature map.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants