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

Is it able to work with perspective image? #7

Open
AtlantixJJ opened this issue Oct 7, 2020 · 7 comments
Open

Is it able to work with perspective image? #7

AtlantixJJ opened this issue Oct 7, 2020 · 7 comments

Comments

@AtlantixJJ
Copy link

In LayoutNetv1 there is a perspective image demo, but I didn't find any relevant script in this repo. Can you help me? Thanks.

@zouchuhang
Copy link
Owner

@AtlantixJJ We do not have the script that runs on perspective images for LayoutNet v2. However, it's straightforward to write the script. You can directly use the LayoutNet v2 training script to train on perspective images with small modifications -- the input perspective image size is 512*512, and skip the alignment and optimization steps, save the output boundary and corner maps into the ".mat" form. Then use the matlab code in the LayoutNet v1 repo to generate the final result.

@AtlantixJJ
Copy link
Author

Thanks. Is it necessary to use matlab? Matlab is hard to configure on server.

@zouchuhang
Copy link
Owner

@AtlantixJJ It's not necessary to use Matlab, it's just because 3 years ago I chose to code it in Matlab . Feel free to rewrite it in Python or C++. I'm happy to see its further improvements.

@ZZWENG
Copy link

ZZWENG commented Oct 15, 2020

alignment

Hi @zouchuhang I am trying to write a perspective image inference script.
Could you specify which part in the train script correspond to the alignment? Also, by "optimization steps", do you mean function optimize_cor_id?
Thanks.

@zouchuhang
Copy link
Owner

@ZZWENG “Could you specify which part in the train script correspond to the alignment?” -- which train script are you referring to? The current LayoutNet v2 training scripts use pre-aligned spherical images as input, so there's no alignment step inside. Yes the optimization refers to the function "optimize_cor_id".

@ZZWENG
Copy link

ZZWENG commented Oct 19, 2020

@ZZWENG “Could you specify which part in the train script correspond to the alignment?” -- which train script are you referring to? The current LayoutNet v2 training scripts use pre-aligned spherical images as input, so there's no alignment step inside. Yes the optimization refers to the function "optimize_cor_id".

Thank you for the clarification!
I re-read the LayoutNet paper and realized that for perspective images, the alignment and manhattan lines are not needed.

@godhj93
Copy link

godhj93 commented Apr 23, 2023

@ZZWENG Hi!, do you have the layoutnetv2 script for perspective images?
if you do can you share it?

thanks :)

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

4 participants