supersyq/RIENet

Questions about experiment

Opened this issue · 2 comments

LiuQiir commented

Thank you for your work, and I hope to learn some knowledge from it. When using your pretrained model for testing, there is a significant discrepancy in the MIE(R) value compared to the data provided in the paper. This issue occurs for all datasets except KITTI. Could you please explain the reason for this?
1d9cd68aede1e8e994087975155a5c1

supersyq commented

I am sorry for my delay. We have tested the pre-trained models on all datasets. The results are similar to those listed in our paper. And there is no significant discrepancy in the MIE(R). Here are the results(7scenes->ICL->same->noise->unseen):
7scenes

  • 👍1
LiuQiir commented

I am sorry for my delay. We have tested the pre-trained models on all datasets. The results are similar to those listed in our paper. And there is no significant discrepancy in the MIE(R). Here are the results(7scenes->ICL->same->noise->unseen): 7scenes

Thank you for your reply, I apologize that I just saw it. I still have a question, as your figures shows, For datasets other than kitti, the metric MIE(R) is much larger than expected. I hope you can help me to answer it!