We mainly use mmdet3d:1.1.x framwork, however, it appears that the visualization hook still not work (lidar-det). So we just make a backup, with two different branchs to train & deploy the same model.
- init mmopenlab 2.0 branch
- train pointpillars in 2.0
- init mmopenlab 1.0 branch
- train pointpillars in 1.0
- visualize the result (mm 1.0)
- import mmdeploy in different branchs