run_360.sh

#!/bin/bash
export CUDA_VISIBLE_DEVICES=1
dataset="nerf_real_360"

position_lr=0.00001
prior_variance=0.01
posterior_lr=0.0001
AUSE_weight=5
KLweight=0.001
datasetargs="--resolution 8"
GSargs="--eval --sh_degree 1 --position_lr_init $position_lr --position_lr_final $position_lr"
bayes_params="--switch_to_beyesian 16000 --prior_xyz_variance $prior_variance --posterior_xyz_lr $posterior_lr --prior_features_variance $prior_variance --posterior_features_lr $posterior_lr --prior_opacity_variance $prior_variance --posterior_opacity_lr $posterior_lr --KL_weight $KLweight --AUSE_weight $AUSE_weight"

for scene in bicycle bonsai counter garden kitchen room stump
do
    input_path=${dataset}/${scene}
    output_path=output/360_${scene}
    python train.py -s $input_path --model_path $output_path $bayes_params $GSargs $datasetargs
    python render.py --model_path $output_path
    python compute_metrics.py --experiment_path ${output_path}/test/ours_30000
done