Awesome

Trimming the Fat

This code is build upon the official implementation of the paper "3D Gaussian Splatting for Real-Time Radiance Field Rendering". To set up the code, please refer to the original repository

For training the baseline model, please use the following command

python train_baseline.py -s <path to COLMAP or NeRF Synthetic dataset> -m <path for the model to be saved>

For training the model with our pruning approach

python train_prune.py -s <path to COLMAP or NeRF Synthetic dataset> -m <path for the model to be saved> --pruning_level <gamma_iter value> --start_checkpoint <path to pre-trained model> --iteration <number of iteration to train>

For evaluating the model

python render.py -m <path to trained model> --iteration <the iteration number at which the model is to be loaded> # Generate renderings
python metrics.py -m <path to trained model> # Compute error metrics on renderings

For evaluating FPS

python fps.py -m <path to trained model> --iteration <the iteration number at which the model is to be loaded> # Calculate FPS