vt-vl-lab / 3d-photo-inpainting

[CVPR 2020] 3D Photography using Context-aware Layered Depth Inpainting
https://shihmengli.github.io/3D-Photo-Inpainting/
Other
6.9k stars 1.11k forks source link

MiDaS v3 #132

Open ugocapeto opened 3 years ago

ugocapeto commented 3 years ago

Amazing project! Thank you for making it available on colab.

I assume it's using MiDaS v2.1. Any plans on making use of MiDaS v3 without boosting (I am really not that convinced boosting makes things better especially with v3) and make it available for us on google colab.

Again, thank you for this.

m0o0scar commented 1 year ago

LOL it's 2023 already still no update on this ... I did some comparson between MiDaS v3.1 and Boost Monocular Depth, MiDaS v3.1 is clearly doing a better job here in terms of correctness and noise level.

Untitled