Introduction
In DeeCamp 2020, our team BUSS developed a virtual LiDAR cloud generation method for data augmentation in 3D object detection pipeline. The point cloud of virtual scenes can be synthesized from manually labeled data through the extraction of different types of objects in real scenes. With the augmentation of training data, the balance of the amount and spatial distribution of different types of objects can be effectively improved.
Pipeline
Results
Videos are available on Youtube!
- Object Collection
- Object Manipulation
- Virtual Cloud Generation
- Visualization
- Final Demo
