Official implementation of our TIV'23 paper: Occupancy-MAE: Self-supervised Pre-training Large-scale LiDAR Point Clouds with Masked Occupancy Autoencoders
Apache License 2.0
251
stars
18
forks
source link
Fixing potential calculation error mentioned in #28 and accelerating shuffling process #32
Fixing potential calculation error mentioned in #28.
Speeds up the shuffling process, in my tests it takes about 50% time of a forward. Modified training speed up by 50%.
However, Please test it before the update.