If 2048 points were used for the experiment, would the results vary a lot

shikiw / SI-Adv

[CVPR 2022] Shape-invariant Adversarial Point Clouds

MIT License

41 stars 10 forks source link

If 2048 points were used for the experiment, would the results vary a lot #5

Open heexu976 opened 1 year ago

shikiw commented 1 year ago

We have not verified it yet. Intuitively, introducing more points can be helpful for the estimation of local normal vectors, thus the performance might be better.