3D-CVF

This is the official implementation of 3D-CVF: Generating Joint Camera and LiDAR Features Using Cross-View Spatial Feature Fusion for 3D Object Detection, built on SECOND.

Requirements

Follow the installation steps in SECOND, or use the docker image we provide.

docker pull yckimm/second:second_v1.5

Getting Started

Training (1st stage)

sh train_bash_1st.sh

Training (2nd stage)

sh train_bash_2nd.sh

Acknowledge

Thanks to the SECOND codebase maintained by traveller59.