Progress for week 48 (2018)
From Robin
Contents |
David Kolden
Budget
- Implement one or more algorithms. Refine D2CO. Research a way to find a gradient with chamfer distance.
Accounting
- Found out that Chamfer Distance Transform can be used as a gradient directly in an IBVC.
- Sketching new algorithms and evaluating them:
- Reducing hypotheses in a IBVC/PBVC hybrid.
- Parallel lines and vanishing points.
- Hybrid configuration set up using model description of gripper together with model description of the object.
- GPU to calculate FDCM?
Jonas S. Waaler
Budget
- Tune data to fit classification task (experiment 2)
- Construct algorithms/architectures to solve the classification task
- First: Traditional methods (similar to experiment 1)
- I suspect I need more data - transfer learning, RNN etc.
Accounting
- Added more metrics
- Pearson Correlation Coefficient
- Liner simularity
- Normalized Cross Correlation
- Simularity with displacement
- MSE of Fast Fourier Transform
- Distance between frequencies
- Pearson Correlation Coefficient
- Added baselines
- Random
- Uniform
- Normal
- Mean of all training data
- Random
Malin Aandahl
Budget
- Implement new fitness function.
- Find out where the position to the robot is calculated(and change to be the back instead of the front of the robot)
- Finish tutorials about Docker.
- If time set up Docker or implement function to find position.