You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Oct 11, 2023. It is now read-only.
Is there a reason why we run ROI detection for training and keep those with large overlap, instead of just using the ground truth bounding boxes with transitional + scaling (random) augmentation?
Is there a reason why we run ROI detection for training and keep those with large overlap, instead of just using the ground truth bounding boxes with transitional + scaling (random) augmentation?