ros package for deep learning based object detection with lidar
This repository is used to deploy trained object detection model in ros ecosystem. It requires the model to be in torchscript format and its output should be Nx7 tensor where N is number of boxes and each box contains class, score, center_x, center_y, width, length, yaw information. To achieve high inference speed, converting pointcloud to voxel is implemented on gpu.
- CUDA
- libtorch
- boost
roslaunch object_detetion dl_detector.launch