Skip to content

Latest commit

 

History

History
20 lines (13 loc) · 925 Bytes

File metadata and controls

20 lines (13 loc) · 925 Bytes

August 2021

tl;dr: Anchor-based method with a ground aware convolution module.

Overall impression

This paper is directly inspired by M3D-RPN. It still uses anchors instead of anchor-free, and uses the postprocessing module to optimize yaw.

Key ideas

  • The key idea is a ground aware convolution (GAC) module. The network predicts the offers in the vertical direction and we sample the corresponding features and depth priors from the pixels below.
  • Depth priors are inspired by CoordConv and are computed with perspective geometry with ground plane assumption.

Technical details

  • Summary of technical details

Notes

  • Questions and notes on how to improve/revise the current work