How to reproduce the performance reported in your paper

[cdfan0627]

您好，想請問一下我按照您的Readme上train nerfds dataset 可是我得到的結果有點差，請問要怎麼做才可以train 出您論文上的結果呢?
這是我下command python train.py -s path/to/your/real-world/dataset -m output/exp-name --eval --is_6dof
以下是我跑出的結果

[ingra14m]

Hi, sry for the late reply.

I think this problem is quite normal since the camera poses in real-world scenes are inaccurate. Therefore, in NeRF-DS dataset, the model will overfit during the training. In my paper, I used the best PSNR, SSIM, and LPIPS evaluated every 1k iteration, not the last metrics after 40k iterations.