Motivation for this project: disease surveillance inspired by robotics/autonomous driving techniques to track the spread of coronavirus.
- Formal input and output of network?
- Input to scene representation graph builder?
- How do they visualize?
- What is the reference frame of each agent? Are these then projected onto a ground plane for visualization purposes?
- Train Trajectron on NuScenes (Divya)
- Write inference code - pytorch (Ching Lam)
- Algorithm to consume output of Trajectron++ network and construct scene representation (graph) (Art)
- Visualization (if time permits)
Deadline: March 16th 5pm PST
- https://arxiv.org/pdf/2001.03093.pdf Trajectron++