Advanced visulization

Advanced visulizations are provided by coperception/tools/visualization/v2xsim_vistoo

The following tutorial and the visualization code are from Yifan Lu

Usage

I first reconstruct the nuscenes-format data, since it's very slow to read? 1. process_v2xsim_v2.py will generate pkl file containing some meta-information of v2x-sim dataset, including lidar pose, gt boxes, and path to actually lidar file.

python dataset_process/process_v2xsim_v2.py

1. folder simple_plot3d is highly from this repo, simple_plot3d/simple_vis.py provide a warpper function for tensor format data. It can draw both BEV image and 3D view image. You can try to integrate the visualize function into your deep learning framework.

2. simple_dataset.py provide a simple dataset design for v2x-sim 2.0, it will retrieve meta information from pkl file and lidar numpy array using lidar file path.

3. Other files like collaboration_view.py, single_view.py, scene_overview.py, location_in_bev.py can draw seqences of pictures that provide a comprehensive overview of the dataset(given a scene).

python single_view.py
python collaboration_view.py
python location_in_bev.py
python scene_overview.py

Then you can use img2video.py to make image sequence to video.

1. You can run visualize_data_seq.py to see how to use the warp function visualize mentioned in 2., just

python visualize_data_seq.py

Result:

GIFs are shown as below: