POTO
Prediction-aware One-To-One
2000년 도입 · 논문 1편에서 사용
Prediction-aware One-To-One, or POTO, is an assignment rule for object detection which dynamically assigns the foreground samples according to the quality of classification and regression simultaneously.
출처: End-to-End Object Detection with Fully Convolutional Network
소개 논문: End-to-End Object Detection with Fully Convolutional Network
Detection Assignment Rules · Computer Vision