New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Investigate possible improvements to tracking performance #258

Open

sfmig opened this issue Nov 21, 2024 · 0 comments

Collaborator

sfmig commented Nov 21, 2024

Why does the detection flicker?

Check detection step only (is it during SORT algorithm?)

Why some "obvious crabs" are missed? (Missed detections)

Try to improve recall
- Longer training
  - Optimise with optima
- Further explore data augmentation
  - Optimise with optuna?
- Refine labelled dataset
  - Review annotations and remove the ones we are not as interested in tracking (right now, we penalise equally the very small ones and the very large ones)

Reduce noise

Sometimes quite a bit of false positives?
Add bboxes sanitation in eval and inference?

Optimise tracking parameters

Run experiments
Consider optuna
Check effect on FP, FN, ID switches separately
Add more ground truth tracks

Robustness tests

See Robustness tests: non-overlapping test/train datasets #178

Optimise

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment