Andréas Meuleman

No it should be the same

Hi, Yes, but without depth and flow losses for now. From initial tests, it also seems to benefit from a higher translation learning rate and skipping frames (if the video...

Hi, Initial experiments on 360 videos seemed to work well. Would you mind sharing the video or a frame? A potential issue that comes to mind is that we often...

Hi, Thanks. The car is a dynamic element that needs to be masked out. Since it covers a large portion of the frame, it hurts pose estimation severely. Luckily, it...

Hi, Cropping the image breaks the model as we are expecting full equirectangular images.

Hi, here are results for each scene. All methods are self-calibrated and the table shows the five scenes the ablation (against `Ours w/o progressive optimization` and `Ours w/o local RF`)...

forest1, forest2, forest3, playground, university1

Hi, Thanks for pointing out this problem, I will take a look.

Hi, A scene optimized with `--with_preprocessed_poses 1 --lr_R_init 0 --lr_t_init 0` can now be rendered with `--render_from_file` properly.

Hi, We preprocess Tanks and Temples scenes as follows: We get the full videos, retain one in every five frames as motion is slow and downscale them to full HD...