Activity
From 16/06/2020 to 15/07/2020
22/06/2020
- 19:55 Product Backlog #15554: Classifier
- Implement classifier, use KNN as that's what was used in the reference paper
- 19:53 Product Backlog #15554 (New): Classifier
- 19:54 Product Backlog #15556 (New): Preprocessing
- Try background subtraction and gaussian model
- 19:53 Product Backlog #15555 (New): Clean data
- 19:50 Feature #15553 (Resolved): Implement KNN classifier
- 19:50 Feature #15553 (Resolved): Implement KNN classifier
- Implement KNN classifier for the optical flow dataset and experiment with different number of neighbors
- 19:49 Bug #15552 (New): Filter optical flow polar coordinates based on magnitude
- Higher magnitude coordinates are probably more important for the classification. Try filtering out low-magniture pola...
- 19:47 Feature #15551: Implement 2D gaussian model to find centroid of the person
- we tried but were unable to make it work reliably
- 19:46 Feature #15551 (Closed): Implement 2D gaussian model to find centroid of the person
- 19:46 Bug #15550 (New): Clean outliers from optical flow output before feeding to classifier
- Optical flow tracing output contains outliers remove them
- 19:44 Feature #15549 (Resolved): Choose between sparse and dense optical flow
- Look into different optical flow algorithm and choose one
- 19:42 Feature #15548 (Resolved): Create background substraction
- 19:42 Feature #15548 (Resolved): Create background substraction
- Look into different ways of creating background subtraction, experiment and choose one that yields best results
Also available in: Atom