The KITTI Vision Benchmark Suite

Our Model Dense-Pointpillars, proposes a dense-
layer based Backbone which is lightweight and has
been designed specifically to improve the
performance for LiDAR-based 3D object detection.
To the best of our knowledge, our paper is the
first to propose dense-layer based lightweight
backbone which is faster, requires less
computations and outperforms the baseline -
Pointpillars. We believe that our paper proposes
an alternative backbone selection choice against
the recently popular sparse convolution-based
backbones and has the potential to propose a
completely new direction of reasearch - building
lightweight backbones to improve performance and
speed of LiDAR-based 3D object detectors.

alpha=0.2