ethnhe / PVN3D

Code for "PVN3D: A Deep Point-wise 3D Keypoints Hough Voting Network for 6DoF Pose Estimation", CVPR 2020
MIT License
482 stars 105 forks source link

Why the size of the pretrained YCB model vary greatly? #81

Closed aiai84 closed 3 years ago

aiai84 commented 3 years ago

I train the YCB dataset and obtain the trained model, about 460M, is the same with Linemod. But the author,s pretrained model is 160M. I have not did any change with the code, what's the reason? Thank you!

aiai84 commented 3 years ago

I have not changed the code and train the YCB dataset, but the result is not as good as the author mentioned in the paper. What is the reason?