I have some questions about evaluation metrics.
What happens if one predicted bounding box has IoU value 0.3 or better with two ground truth bounding boxes? And if one ground-truth bounding box intersected with high IoU with several predicted bounding boxes?
Is it possible to share a evaluation metrics soure code to make sure that we understand it correctly?
Good question. If one ground truth box has an IoU value of 0.3 or better with two or more truth bounding boxes, then we choose only one corresponding ground truth bounding box with the highest IoU value.
1 Like