Table 3 Tunable hyper-parameters elaboration for the paragraphs of prediction post-processing and “bagging” in Sect. 2.2.2.
Hyper-parameters | Explanation | Tuning range | Section |
|---|---|---|---|
τ1 | IoU threshold for prediction filtering | [0.1, 0.2, 0.3, …, 1.0] | Window bagging |
τ2 | Threshold for including final voter numbers | [1–5] | Window bagging |
τ3 | IoU threshold for determination of bagged prediction hitting GTs | [0.1, 0.2, 0.3, …, 1.0] | Window bagging |
τ4 | Threshold for filtering predicted detection box in expansion zone | [1, 2, 3, …, 10] | Post-processing |
τ5 | Absolute threshold for filtering HU | [80, 100, …, 160] | Post-processing |
τ6 | Quantile threshold for filtering HU | [0.1, 0.2, 0.3, …, 1.0] | Post-processing |
