question about TNT dataset performance

[XipengY]

Hi, very interesting work！
We cannot reproduce the metrics of the paper in TNT dataset, the train and test scripts：
`
#training

python train.py -s data/TNT_GOF/TrainingSet/Truck -m output/Truck -r 2 --eval --use_decoupled_appearance

#mesh extraction

python mesh_extract_tetrahedra.py -s data/TNT_GOF/TrainingSet/Truck -m output/Truck -r 2 --eval

#evaluation

python eval_tnt/run.py --dataset-dir data/GT_TNT_dataset/Truck --traj-path data/GT_TNT_dataset/Truck/Truck_COLMAP_SfM.log --ply-path output/Truck/recon.ply
`
The train dataset is download from GOF, both the train and evaluation dataset are followed by README.

we evaluation the scene of Truck: [The f1-score in paper is 0.53]

What is this problem? Can you help us.