RT3D: Real-Time 3D Vehicle Detection in LiDAR Point Cloud for Autonomous Driving

This is repositoy of paper "RT3D: Real-Time 3D Vehicle Detection in LiDAR Point Cloud for Autonomous Driving"

Our implementation are based on the RFCN by YuwenXiong: https://github.com/YuwenXiong/py-R-FCN

And the Caffe should use the Mircosoft brunch :https://github.com/Microsoft/caffe.git

Runtime support: CUDA 8.0, CUDNN-v5

The results on validation set is here: download

ResNet model

$ROOT/data/imagenet_models/ResNet-50-deploy.prototxt

$ROOT/data/imagenet_models/ResNet-50-model.caffemodel

Data

Kitti object detection data should transfer to Pascal VOC style.

For example: $ROOT/data/VOCdevkit2007

Running

To runnung testing

$ python test_theta_w_l.py --gpu id --img BEVPATH