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

Depth Evaluation Alignment #33

Open
Davidyao99 opened this issue Oct 28, 2024 · 1 comment
Open

Depth Evaluation Alignment #33

Davidyao99 opened this issue Oct 28, 2024 · 1 comment

Comments

@Davidyao99
Copy link

Hi, I was wondering if you performed the same alignment on all the different depth predictions during your evaluations? I tried to apply the same alignment (lad2) to other prediction results such as depthcrafter but it leads to some outlier values (extremely large values in some pixels) that leads to strange results in metrics such as video abs rel. I was wondering if you have experienced this in your experiments.

Additionally, given that the metrics seem to differ using different alignment methods, I was wondering if you have any insights on how to pick the best alignment methods during evaluation.

Thank you so much for your help! Much Appreciated!

@Junyi42
Copy link
Owner

Junyi42 commented Oct 29, 2024

Hi @Davidyao99,

We applied the optimal depth alignment approach for each method. And that's why we are applying lad2 for our method and lstsq for DepthCrafter. On finding the better alignment approaches, we will first try a few traditional ways, and will use the one that gives overall better results across different metrics.

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

2 participants