fdarmon/NeuralWarp

The result on DTU scan24 does not align to the paper

SAOHPRWHG opened this issue · 3 comments

Follow the instruction of this repo. I got a result, 0.72, on DTU scan24 scene, which is different from the result in this paper, 0.49.
Could you please tell me if anywhere I made some mistakes?
image

These are my training and extracting commands:
python train.py --conf confs/NeuralWarp.conf --scene 24
python extract_mesh.py --conf confs/NeuralWarp.conf --scene 24

It looks like you have not trained the baseline model before finetuning?

First run python train.py --conf confs/baseline.conf --scene 24 then python train.py --conf confs/NeuralWarp.conf --scene 24

Thanks! I got the correct result.