Produced result is a little bit worse

[skhu101]

I use the following command, but it seems that the performance is worse than yours. May I ask do I miss any things?

CUDA_VISIBLE_DEVICES=1 python train.py
--root_dir data/DTU/mvs_training/dtu
--num_epochs 6 --batch_size 1
--n_depths 192 --interval_scale 1.06
--optimizer adam --lr 1e-3 --lr_scheduler cosine