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
Hi, thanks for your work. I have a question about the design. SortedAP is calculated as the area under the IOU-AP curve. However, the absolute value of the IOU threshold is not taken into account, only the IOU difference between points is used. So if the AP is the same for a low IOU threshold point and a high IOU threshold point, they will contribute the same to the sortedAP calculation. Intuitively, the higher IOU threshold value should be more important. So maybe when calculating, we can weight the summation by the absolute IOU value?
The text was updated successfully, but these errors were encountered:
Hi @menggui1993 , thank you for your question. Actually, the absolute IOU is considered. A higher object IOU will lead to a decrease in the AP curve at a higher IOU value, so the area under the curve will be larger than the case of a lower object IOU. You can refer to Figure 2b in the paper, in which we both have 3 predictions but with different IOUs.
Hi @menggui1993 , thank you for your question. Actually, the absolute IOU is considered. A higher object IOU will lead to a decrease in the AP curve at a higher IOU value, so the area under the curve will be larger than the case of a lower object IOU. You can refer to Figure 2b in the paper, in which we both have 3 predictions but with different IOUs.
Hi, thanks for your work. I have a question about the design. SortedAP is calculated as the area under the IOU-AP curve. However, the absolute value of the IOU threshold is not taken into account, only the IOU difference between points is used. So if the AP is the same for a low IOU threshold point and a high IOU threshold point, they will contribute the same to the sortedAP calculation. Intuitively, the higher IOU threshold value should be more important. So maybe when calculating, we can weight the summation by the absolute IOU value?
The text was updated successfully, but these errors were encountered: