search

tsinghua-rll / VoxelNet-tensorflow

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

Applying mask in the VFELayer #28

Closed rmeertens closed 6 years ago

rmeertens commented 6 years ago

When profiling the runtime I noticed that the tiling of the mask to mask = tf.tile(mask, [1, 1, 2 * self.units]) takes a long time. I was wondering why this mask is applied, don't empty voxels already get filtered out when selecting the pointcloud? Would love to know if there is any reason behind this code...

jeasinema commented 6 years ago

Hi,

This line of code has noting to do with (non-)empty voxels, but redundant points in one voxel. Since some voxels may contain less than T points, we just augment them to T. After the multiply operation, values coming from such augmented points need to be fixed at zero.