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

Inclination is incorrect #3

Open
Mps24-7uk opened this issue Sep 21, 2021 · 9 comments
Open

Inclination is incorrect #3

Mps24-7uk opened this issue Sep 21, 2021 · 9 comments
Labels
question Further information is requested

Comments

@Mps24-7uk
Copy link

Thanks for the amazing work. I tried the running test.py on nyu dataset and blended the seg to check the alignment. It seems to be going off. Can you suggest how to fix it? Please find the example below.

Blended segmentation and image
9

segmentation and image
9

@bertjiazheng
Copy link
Collaborator

Hi,

Thanks for your interest in our work. I am not exactly sure what happens, but you could check the 2D detected lines are correct or not.

Best,
Jia

@Mps24-7uk
Copy link
Author

@bertjiazheng I checked the 2D lines, getting it wrong. Could you please help, check it once why the output layout is not inclined.

@CYang0515
Copy link
Owner

What the other cases in the NYU303 dataset?

@Mps24-7uk
Copy link
Author

Mps24-7uk commented Oct 14, 2021

@CYang0515 I tried out on NYU303 as well as custom test set but it didnt worked for me. You have given the padding of (480, 640, 3) to the input image of size (427,561,3) . I did the post processing but it didn't fixed the issue.

@CYang0515
Copy link
Owner

What resolution of your origin image in NYU303 dataset? Maybe the images you used are not cropped and resized, as the processed images do not have white border.

@Mps24-7uk
Copy link
Author

Mps24-7uk commented Oct 14, 2021

In nyu303 dataset the fullres image size is (480,640,3) and cropped image size is (427,561,3). I checked on nyu303 images and the result is doesn't have correct alignment. In the result you will see that the layout is not exactly aligned with the wall edges

37

97

@CYang0515
Copy link
Owner

We add one case of NYU dataset for debug, you can type python test.py --pretrained DIR --data NYU303 --exam --visual to get right result in ./results.

@CYang0515
Copy link
Owner

We have updated the code, and provide an --exam parameter to show one example of NYU303 dataset. The input image and reconstruction results are in example/NYU0969 and results separately.

@Mps24-7uk
Copy link
Author

Mps24-7uk commented Oct 15, 2021

Thanks for updating the repo. The NYU303 test set results look promising but it fails in some cases. In the result, the predicted wall is not correct. Is it because we are predicting the layout on the crop image and trying to fit it on fullres image.

40

45

Can we use --pretrained NYU303_pretrained.pt --data NYU303 this setting to estimate the layout for custom test image?

@bertjiazheng bertjiazheng added the question Further information is requested label Oct 21, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
question Further information is requested
Projects
None yet
Development

No branches or pull requests

3 participants