Pretrained TUM RGBD and tips on implementing per-frame

[Any0019]

Hello, thank you for such a great work. Was very interested in two things for my own project.

1. Do you have an opened pretrained model for TUM-RGBD, which you have shown in the demo?
2. I want to use your approach in my own work for per-frame localization function in close-space environment and was wondering is there any easy way to use your model the way that you will give it position on step t (last position) and new frame (after moving) on every step and it will return you updated position w.r.t. to new image and the actual move that has been done?
   Would be very grateful for your help.