tsinghua-rll / VoxelNet-tensorflow

A 3D object detection system for autonomous driving.
MIT License
453 stars 123 forks source link

Speed up VFE layer #5

Closed ring00 closed 6 years ago

ring00 commented 6 years ago

This commit partially solves issue #3.

However, according to the paper, the features corresponding to empty points should be reset to zero after concatenation operations. Current implementation uses a boolean mask to achieve this, but this method is not completely correct.

@jeasinema I didn't test the code so there may be bugs and you may want to do some minor modifications.