Congrong Xu
Congrong Xu
I think this is kind of tricky. The whole pipeline is instantiated when you run viser, and all the images is cached when the datamanager is instantiated with the pipeline....
Wondering if i should seperate get_outputs_for_camera from get_outputs to git rid of lots of if self.training. And viser currently overwrite bg color in set_background() thus requiring resetting the training/rendering bg...
> The PR looks almost good to me (except one minor comment), but it will need to wait until the next gsplat version update before merging this PR (maybe in...
 Here's the bicycle scene predict 2d bg instead
> My general thoughts on this: allowing background to be learnable should be used with some regularizations (e.g. minimizing L1 norm of accumulation map etc.) Otherwise this still is uncontrollable....
I think the easiest way is to preprocess the imgs and remove their backgrounds. https://github.com/danielgatis/rembg may help.
maybe try add prompt "full body"
Hi! Made a pr for this in https://github.com/nerfstudio-project/nerfstudio/pull/3146.
> @KevinXu02 The PR will be merged when we produced a new version of gsplat library that support background parameter grad. Got it. Thank you!
> > torch.tensor(camera["camera_to_world"]).view(4, 4)[:3] > > I found is my "camera_path": []", camera_path is empty, but what "camera_path" means ??? Hi! This should be the camera path to render your...