Online Demo question.

[DeepHM]

What is SfMNeXt-indoor(78M), SfMNeXt-outdoor(78M), SfMNeXt-outdoor-cvnxt(242M), ZoeDepth(345M), ZoeDev and MidaS(345M) in model type of online demo?

[hisfog]

These model types differ in model architecture, training data, training strategies, I'm not sure which aspect you would like to know about?

[DeepHM]

Thank you for your quick reply.

Here are my additional questions :

1. On the "paper with codes" site, I saw this git-repository linked to the paper "SQLdepth: Generalizable Self-Supervised Fine-Structured Monocular Depth Estimation". Is this repository an implementation of the SQLdepth paper?
2. If question 1 is correct, is SfMNeXt the model for the SQLdepth paper?
3. I would like to know the training data used for the models SfM NeXt-indoor(78M), SfM NeXt-outdoor(78M), SfM NeXt-outdoor-cvnxt(242M), ZoeDepth(345M), ZoeDev, and MidaS(345M). If you know, please let me know.

[hisfog]

1. Yes, but more accurately, this repository includes implementations of the paper you mentioned, but it goes beyond that. In fact, this repository will encompass all of my research work related to self-supervised depth estimation.
2. Yes, except for SfMNeXt-indoor.
3. SfMNeXt-indoor is self-supervised pretrained on MC dataset, and finetuned on NYU-Depth Dataset. SfMNeXt-outdoor(78M) is self-supervised trained on KITTI, SfMNeXt-outdoor(242M) is self-supervised pretrained on KITTI, and finetuned on KITTI. ZoeDev is self-supervised finetuned on MC dataset. As for ZoeDepth and MiDaS, please refer to their paper.

[DeepHM]

Thank you!!!