search

ZrrSkywalker / I2P-MAE

[CVPR 2023] Learning 3D Representations from 2D Pre-trained Models via Image-to-Point Masked Autoencoders
219 stars 17 forks source link

Modelnet40 #11

Open sehunfromdaegu opened 7 months ago

sehunfromdaegu commented 7 months ago

Hi, thanks for sharing the nice work.

I wonder why do you use two different versions of Modelnet40 data for fine-tuning, and linear SVM. Is there any benefit of this?