image

paper

TL;DR

  • task : uncertainty in FCOS model
  • problem : left, right, top, bottom 4 방향에 대한 uncertainty를 얻고 싶다.
  • idea : 4개의 가우시안을 가정하자
  • architecture : FCOS + uncertainty branch + classification head에 uncertainty branch의 output을 fusion
  • objective : uncertainty loss : 4개의 가우시안에 대한 KL divergence에 IoU term을 지수항으로 넣어줌. cls loss : uncertainty를 반영한 focal loss
  • baseline : centerness-branch, IoU branch, QFL, VFL
  • data : COCO2017
  • result : FCOS 성능 1.8 point 올림

Details

image

Architecture

image

Loss

Uncertainty Loss

image

Uncertainty Focal Loss

image

Result

image image image