This adapts gaussian splats to be used on amateur(sparse, shaky, blurry) iphone data.
This adds:
- camera pose learning
- time variable
- deblur(https://github.com/Chaphlagical/Deblur-GS)
- importance sampling[Learns the frame and patches with the most intersting features across time and spacetime]
This started as a fork from:
Deblur-GS: 3D Gaussian Splatting from Camera Motion Blurred Images Official implementation of paper "Deblur-GS: 3D Gaussian Splatting from Camera Motion Blurred Images", I3D 2024
SET DISTUTILS_USE_SDK=1 # Windows only
conda env create --file environment.yml
conda activate deblur_gs
python train.py -s <path to dataset> --eval # Train with train/test split
Additional Command Line Arguments for train.py
blur_sample_num
: number of key frames for trajectory time samplingdeblur
: switch the deblur modemode
: models of camera motion trajectory (i.e. Linear, Spline, Bezier)bezier_order
: order of the Bézier curve when use Bézier curve for trajectory modeling
python train.py -s <path to dataset> --eval # Train with train/test split
python render.py -m <path to trained model> # Generate renderings
python metrics.py -m <path to trained model> # Compute error metrics on renderings
Additional Command Line Arguments for render.py
optim_pose
: optimize the camera pose to align with the dataset
python render_video.py -m <path to trained model>
@article{Chen_deblurgs2024,
author = {Wenbo, Chen and Ligang, Liu},
title = {Deblur-GS: 3D Gaussian Splatting from Camera Motion Blurred Images},
journal = {Proc. ACM Comput. Graph. Interact. Tech. (Proceedings of I3D 2024)},
year = {2024},
volume = {7},
number = {1},
numpages = {13},
location = {Philadelphia, PA, USA},
url = {http://doi.acm.org/10.1145/3651301},
doi = {10.1145/3651301},
publisher = {ACM Press},
address = {New York, NY, USA},
}