The following repository provides the open source implementation of complex-yolo for 3d object detection in BEV map of point cloud.
But he problem is there while training, loss is not conversing and those third party implementation seems to be incomplete. Here we tried to complete the project based on above implementations.
https://github.com/ghimiredhikura/Complex-YOLO
TODO:
- Add anchor calculation script - completed
- data argumentation
- training
- evaulation