Skip to content

YoungYoung619/bev-perception-learning-for-autonoumous-driving

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

bev-perception-learning-for-autonoumous-driving

使用nuscenes数据集,复现了一些bev算法,仅供学习

Lift Splat Shoot

Train

以下脚本用于开启训练,请在对应配置文件中指定好您的nusences数据集位置和您的训练配置(gpu数量等)

export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
cfg_file=./config/lift_splat_shoot/lss_segm.yml
python tools/train.py --config ${cfg_file}

Demo

以下脚本用于可视化模型检测效果,请在对应配置文件中指定好您的数据集位置,然后您将会看到如图1.所示的检测效果,其中,左侧六张图像为camera输入,右侧上方为真值,右侧下方为网络预测值,预训练的网络可以在model zoo中下载 (LSS & Segm)

python demo/demo_segm.py --cfg_file=${cfg_file} --model_file=${model_ckpt} --device=mps
图.1 Lift splat shoot 分割检测头效果

Bevdepth

Train

export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
cfg_file=./config/bevdepth/centerhead_3ddet.yml
python tools/train.py --config ${cfg_file}

Demo

python demo/demo_3ddet.py --cfg_file=${cfg_file} --model_file=${model_ckpt} --device=mps
图.2 Bevdepth 3d目标检测头效果

Model Zoo

以下为模型预训练文件及训练log,验证集指标可以在log中查看

Method Type Model Log
LSS Segm Google Google
bevdepth 3D检测 Google Google

About

复现了一些bev算法,仅供学习

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages