This repository provides the implementation of the methods described in the project "Sistema de seguimiento de pose de cajas para asistir operaciones manuales de cubicaje de mercancía: enfoque de integración de SLAM RGBD con conocimiento previo de secuencia de cubicaje, cantidad y tamaño de cajas en la escena". The system processes RGBD point clouds from a head-mounted camera, identifies plane segments, and tracks 3D positions and sizes of box-shaped objects over time. The associated papers are
- Camacho-Muñoz, G. A., Franco, J. C. M., Nope-Rodríguez, S. E., Loaiza-Correa, H., Gil-Parga, S., & Álvarez-Martínez, D. (2023). 6D- ViCuT: Six degree-of-freedom visual cuboid tracking dataset for manual packing of cargo in warehouses. Data in Brief, 109385. https://doi.org/10.1016/j.dib.2023.109385
2.Camacho-Muñoz, G.A., Nope Rodríguez, S.E., Loaiza-Correa, H. et al. Evaluation of the use of box size priors for 6D plane segment tracking from point clouds with applications in cargo packing. J Image Video Proc. 2024, 17 (2024). https://doi.org/10.1186/s13640- 024-00636-1
- Camacho-Muñoz, G.A., Nope Rodríguez, S.E., Loaiza-Correa, H. Evaluating 6D Box Tracking in Real-World Packing Operations: Effects of Prior Knowledge and Camera Velocity.
- RGBD SLAM-based point cloud acquisition (HoloLens 2 compatible).
- Local and global plane segment extraction.
- Plane segment clustering and box reconstruction.
- Pose and size estimation with prior-based adjustment.
- Performance metrics computation (Precision, Recall, F1-score, error metrics).
Watch the demo video showing the pipeline execution step by step:
- Reading Input Point Cloud
- Generation of Local Plane Segments
- Generation of Map of Local Plane Segments
- Generation of Global Plane Segment Map
- Generation of Global Box Map
