The question about simulation displacement.

[newstudy886]

Hi, thank you for yours for providing the code and the evaluation metrics on your GitHub repository.
When I reproduced the work of your grabnet, the other metrics were similar. However, the simulation displacement was quite different from the original article. The average simulation offset in the paper was 1.21cm but the running result was 2.04cm. What are the possible reasons?
I would be very grateful if I could get your reply.

[lixiny]

Hi, did you try our checkpoints on these metrics?

[newstudy886]

Thanks for your reply. Yes, I use your public code and checkpoints without any changes. Multiple experiments have found that there is a certain difference in this value, fluctuating around 2.00cm. In my opinion, it may be due to the randomness of the generation problem itself, and each generation result has its own differences.