where's thresholds.npy come from?

[QiaoranC]

Hi, i found you using different thresholds for each class, this seems interesting. How did you generate the thresholds for each class? and why?

[thtang]

The thresholds are set to maximize the sum of True positive rate and True negative rate for each class on the validation set. In other words, sensitivity and specificity are considered here.

[QiaoranC]

Emm, not fully understand why, but thanks for the explain

[QiaoranC]

i see this iou.oy and find_bbox_size.py, this are more like to find the best npy from a list npy. how do i generate the thresholds.npy? Thx

[thtang]

When doing validation, compute ROC curve for each class. sklearn.metrics.roc_curve could help you. You would get three returns, fpr, tpr and thresholds. Then find out the threshold that could maximize (tpr + (1-fpr)) for each class.

[QiaoranC]

got it thx again