Hi !
I try to run Mega-NeRF on KITTI dataset, sequence 00. It's a dataset, containing images from stereo camera, facing forward, on a car. I took only images from left camera from stereo camera.
At first I tried to train first 20 frames of the sequence without splitting into cells, i.e. I defined grid_dim 1 1 when running script create_cluster_masks.py.
In config file I put only one parameter:
ray_altitude_range: [-200, +200]
I expect that rendering ray will not be farther than 200 meters from the initial pose of the car in the sequence of images even after 20 frames (the car will not go away more than 20 meters from the starting point).
After training on 500 000 iterations I've run merge_submodules.py and create_octree.py.
Then I opened results in mega-nerf-viewer and saw something unreasonable - see the attached screenshot
Hi ! I try to run Mega-NeRF on KITTI dataset, sequence 00. It's a dataset, containing images from stereo camera, facing forward, on a car. I took only images from left camera from stereo camera. At first I tried to train first 20 frames of the sequence without splitting into cells, i.e. I defined grid_dim 1 1 when running script
create_cluster_masks.py
. In config file I put only one parameter:ray_altitude_range: [-200, +200]
I expect that rendering ray will not be farther than 200 meters from the initial pose of the car in the sequence of images even after 20 frames (the car will not go away more than 20 meters from the starting point). After training on 500 000 iterations I've runmerge_submodules.py
andcreate_octree.py
. Then I opened results in mega-nerf-viewer and saw something unreasonable - see the attached screenshotWhat have I done wrong ?