Thank you for your excellent work. I ran the code you provided, but the result was lower than what was reported in your paper. Can you provide me with the correct running parameters
run EO-NeRF by Roger Marí
full paper: "Multi-Date Earth Observation NeRF: The Detail Is in the Shadows" (CVPR Workshops 2023)
timestamp=$(date +"%Y-%m-%d_%H-%M-%S")
hyperparams and key vars
aoi_id=JAX_260
suffix=$2
dataset=JAX_RGB
gpu_id=0
downsample_factor=2
n_samples=128
n_importance=0
fc_units=256
training_iters=300000
batch_size=1024
input paths
datasetdir=/home/lf/PycharmProjects/eonerf_code-main/Datasets
root_dir=$datasetdir"/SatNeRF/root_dir/crops_rpcs_ba_v2/"$aoi_id
cache_dir=$datasetdir"/SatNeRF/cache_dir_utm/crops_rpcs_ba_v2/"$aoi_id"_ds"$downsample_factor
img_dir=$datasetdir"/DFC2019/Track3-RGB-crops/$aoi_id"
gt_dir=$datasetdir"/DFC2019/Track3-Truth"
shadow_masks_dir=$datasetdir"/DFC2019/Shadows-pred_v2/Track3-RGB-crops/"$aoi_id
output paths
out_dir="/home/lf/PycharmProjects/eonerf_code-main/output"
logs_dir=$out_dir/logs
ckpts_dir=$out_dir/ckpts
errs_dir=$out_dir/errs
mkdir -p $errs_dir
errs="$aoi_id"_errors.txt
input arguments
model="eo-nerf"
exp_name="$timestamp""$aoi_id""$dataset"_eonerf
extra_args="--model $model --img_downscale $downsample_factor --max_train_steps $training_iters --fc_units $fc_units --n_samples $n_samples --n_importance $n_importance --geometric_shadows --radiometric_normalization --batch_size $batch_size"
add shadow supervision and use only 9 views
#exp_name="$exp_name"_9views_shadowsupervision
#extra_args="$extra_args --subset_Nviews 9 --shadow_masks_dir $shadow_masks_dir"
run model
errs=$errs_dir/"$exp_name"_errors.txt
echo -n "" > $errs
python3 train_eonerf.py --exp_name $exp_name --root_dir $root_dir --img_dir $img_dir --cache_dir $cache_dir --ckpts_dir $ckpts_dir --logs_dir $logs_dir --gt_dir $gt_dir --gpu_id $gpu_id $extra_args #2>> $errs
Thank you for your excellent work. I ran the code you provided, but the result was lower than what was reported in your paper. Can you provide me with the correct running parameters
run EO-NeRF by Roger Marí
full paper: "Multi-Date Earth Observation NeRF: The Detail Is in the Shadows" (CVPR Workshops 2023)
timestamp=$(date +"%Y-%m-%d_%H-%M-%S")
hyperparams and key vars
aoi_id=JAX_260
suffix=$2
dataset=JAX_RGB
gpu_id=0
downsample_factor=2
n_samples=128
n_importance=0
fc_units=256
training_iters=300000
batch_size=1024
input paths
datasetdir=/home/lf/PycharmProjects/eonerf_code-main/Datasets
root_dir=$datasetdir"/SatNeRF/root_dir/crops_rpcs_ba_v2/"$aoi_id
cache_dir=$datasetdir"/SatNeRF/cache_dir_utm/crops_rpcs_ba_v2/"$aoi_id"_ds"$downsample_factor
img_dir=$datasetdir"/DFC2019/Track3-RGB-crops/$aoi_id"
gt_dir=$datasetdir"/DFC2019/Track3-Truth"
shadow_masks_dir=$datasetdir"/DFC2019/Shadows-pred_v2/Track3-RGB-crops/"$aoi_id
output paths
out_dir="/home/lf/PycharmProjects/eonerf_code-main/output"
logs_dir=$out_dir/logs
ckpts_dir=$out_dir/ckpts
errs_dir=$out_dir/errs
mkdir -p $errs_dir
errs="$aoi_id"_errors.txt
input arguments
model="eo-nerf"
exp_name="$timestamp""$aoi_id""$dataset"_eonerf
extra_args="--model $model --img_downscale $downsample_factor --max_train_steps $training_iters --fc_units $fc_units --n_samples $n_samples --n_importance $n_importance --geometric_shadows --radiometric_normalization --batch_size $batch_size"
add shadow supervision and use only 9 views
#exp_name="$exp_name"_9views_shadowsupervision
#extra_args="$extra_args --subset_Nviews 9 --shadow_masks_dir $shadow_masks_dir"
run model
errs=$errs_dir/"$exp_name"_errors.txt
echo -n "" > $errs
python3 train_eonerf.py --exp_name $exp_name --root_dir $root_dir --img_dir $img_dir --cache_dir $cache_dir --ckpts_dir $ckpts_dir --logs_dir $logs_dir --gt_dir $gt_dir --gpu_id $gpu_id $extra_args #2>> $errs