1. Quick Description

1.1 Dependency Graph

We use Dependency Graph to connect components to avoid redundancy.

Where four building blocks are involved in the current factory, (1) Mono-SLAM (Our extended-DPVO, supports RGB, RGB-D, RGB-L), (2) depth estimation (Our U2-MVD), (3) Flexion and (4) scene reconstruction (Our DM-NPs).

---

1.2 Due-purpose Multiresolutional Neural Points (DM-NPs)

1.2.1 The first surface-gettable SLF model.

1.2.2 Make point rasterization as usable as surface rasterization with improved point rasterization (IPR).

The Pytorch3d's result also uses our layer filter for variables control.

.

1.3 Unposed&Uncalibrated Multi-view Depth Estimation (U2-MVD)

Our model doesnot suffers from the vision illusion.

1.4 Fill in the graph