I got diifferent results on the evaluate set.

[BinBCheng]

Hello, thanks for your wonderful work. I got the pretrained model and I ran the train.py to reproduce the dynamic-multiframe-depth, but I got different results from yours(resnet18-pretrained).

Paper Abs_rel / Sq_rel / rmse / rmse_log / a1 / a2 / a3 0.043 0.151 2.113 0.073 0.975 0.996 0.999

Own Abs_rel / Sq_rel / rmse / rmse_log / a1 / a2 / a3 0.126 0.893 4.552 0.19 0.833 0.94 0.981

And my torch==1.10.1+cu113,torchvision==0.11.2+cu113.

All indicators are quite different from those in the paper. I did not make changes in “trian_my_resnet18.json”, just replaced “n_gpus=8” with “n_gpus=3”.

1. I want to know why my result is not good.
2. In addition to the settings in “trian_my_resnet18.json”, what details do I need to pay attention to in order to reproduce the result.

Looking forward to your reply, thank you. Best wishes!

[ruili3]

Hi, Thanks for your attention to our work. The results look weird, can you double-check if the scores are from dynamic area evaluation or full-image evaluation?