A 3D view the DeepTraffic project as part of MIT 6.S094 Deep Learning for Self-Driving Cars
https://chrmi.github.io/DeepTraffic3D/
This project aims to give a 3D view to the DeepTraffic project using Unity, targeting WebGL (as to leverage all of the same JavaScript functionality of the existing project).
https://selfdrivingcars.mit.edu/deeptraffic/
1.17.2018:
- Simple first-person view from an Asset Store Collada car model.
- Animated textures on the road to give a motion implication (same as the falling dotted lines in the original).
- Animated sunset
- Several PoC cars are introduced to the scene on load.
- Cars are propelled forward using physics (force), vs fixed interval -- this will allow for applying a gas pedal, vs moving fixed units as speed.
- WebGL Template including auto-resizing canvas (responsive), and integration of ConvNetJS.
- Top-down Camera (disabled, used for debugging).
1.17.2018:
- There is cross communication between the Unity and Browser layers to facilitate loading, but this should be expanded to support feeding ConvNetJS back into Unity.
- Better 3D models / graphics / design.
- A means to load simulations
- Support for simulations (remove placeholder cars).
- UI layer for switching between first person and top-down camera.
- Too much to list -- this is just an initial proof of concept.