I need to get my thesis proposals done and shop them around to potential supervisors.
- Finish topics: Active Perception, Stochastic Optimization
- Try the disparity filtering thing
Finish this thing and see if it’s worth a paper / open sourcing it.
- Investigating making the Kalman Filter dynamic, using # of points, maybe viewing angle
- Extend to moving vehicles w/car motion model from Aachen paper, figure out better tracking
- Try sampling points adaptively (i.e. maintain consistency like in LSD)
- Try some hacks to see if we can make this faster
- Get KITTI laser data to give quantitative result
Get this written up with some demo code and posted here.
- Run LSD SLAM in an isolated environment
- Record data
- Write up fault, causes, possible solutions
Chameleon tracker is an interesting paper, and I think I can extend the author’s ideas.
- Annotate and summarize paper
- Attempt implementation
Human Intent Prediction
Need to see if there is a predictive signal in the data we collected.
- Visualize trajectories VS human pose
- Attempt basic statistical analysis
Working 2 days this week.