boostcampaitech3 / level2-data-annotation_cv-level2-cv-09

level2-data-annotation_cv-level2-cv-09 created by GitHub Classroom
0 stars 4 forks source link

[Discussion] Predicted Box 2 Ground Truth Ratio #2

Closed yoonghee closed 2 years ago

yoonghee commented 2 years ago

What?

Why?

가정) Detection Result가 Ground Truth Region의 2배일 때 가장 성적이 잘 나오지 않을까?

One-to-many의 경우 recall / precision이 0.8로 패널티 적용

TO DO