Thank you for your hard work on this project. I have a question regarding depth estimation: how does your model learn the depth of the sky in the images? Since there is no ground truth depth for the sky, do you assign a specific depth value to these areas, or do you simply ignore them?
In my understanding, if you ignore the sky area during the training process, the inference results could be affected, and the model might randomly predict a near depth for the sky, since it never learns how to estimate the depth of the sky correctly.
I also noticed that you provided an inferred image where the depth of the sky seems correct. Could you please explain how you achieve this?
lizhiqi49
commented
4 days ago
Hi,
Thank you for your hard work on this project. I have a question regarding depth estimation: how does your model learn the depth of the sky in the images? Since there is no ground truth depth for the sky, do you assign a specific depth value to these areas, or do you simply ignore them?
In my understanding, if you ignore the sky area during the training process, the inference results could be affected, and the model might randomly predict a near depth for the sky, since it never learns how to estimate the depth of the sky correctly.
I also noticed that you provided an inferred image where the depth of the sky seems correct. Could you please explain how you achieve this?
Thank you for your assistance.
Best regards