Giters

SamsungLabs / imvoxelnet

[WACV2022] ImVoxelNet: Image to Voxels Projection for Monocular and Multi-View General-Purpose 3D Object Detection

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Can model inference without "lidar2img" matrix?

ChenSi521 opened this issue · comments

commented

I have read some your model code, but when I want to inferece by my image, I find I can't create a 'lidar2img' matrix。Can you tell how to inference without 'lidar2img'？

commented

Hi @ChenSi521,

You are right, that for almost all our datasets we assume known intrinsic and extrinsic camera matrices. This is not suprising for outdoor datasets as we know all about sensors at the time they are placed on the vehicle.

The only model that can be run without lidar2img is for SUN RGB-D Total3DUnderstanding benchmark. Here we predict camera angles with additional 2d head.