search

charlesq34 / pointnet

PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation
Other
4.76k stars 1.45k forks source link

Question about FLOPs/sample #216

Closed Xiangxu-0103 closed 4 years ago

Xiangxu-0103 commented 4 years ago

Hi I'm trying to calculate the FLOPs/sample you mentioned in the paper, but it has a large difference with the result you reported. Here is my code! Wish your reply! 1

WuzhaoLee commented 4 years ago

Hello, I have this question, too. I use the same code as yours, then I calculate the FLOPs/sample as: 899521486, (about twice the results in the paper) and params as 3480049. (The model is 'pointnet_cls.py'). What about your calculation results?

Xiangxu-0103 commented 4 years ago

Hi: I'm so sorry to respond to your message so long. The FLOPs reported in PointNet is float op per SAMPLE and didn't include BN etc. op. Best wishes. Xiang.

WuzhaoLi notifications@github.com 于2020年1月3日周五 下午4:47写道：

Hello, I have this question, too. I use the same code as yours, then I calculate the FLOPs/sample as: 899521486, (about twice the results in the paper) and params as 3480049. (The model is 'pointnet_cls.py'). What about your calculation results?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/charlesq34/pointnet/issues/216?email_source=notifications&email_token=ANHATYXIFTWCCPYTUZRGX23Q333TRA5CNFSM4JHX3S62YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEIATUWA#issuecomment-570505816, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANHATYRJDODI5HAHE6ZK4CDQ333TRANCNFSM4JHX3S6Q .