IoU-balanced classification and localization losses, which up-weight high-IoU examples and down-weight low-IoU examples during training, improve AP by about 1.0 to 1.7 points on COCO test-dev for several single-stage detectors.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
IoU-balanced Loss Functions for Single-stage Object Detection
IoU-balanced classification and localization losses, which up-weight high-IoU examples and down-weight low-IoU examples during training, improve AP by about 1.0 to 1.7 points on COCO test-dev for several single-stage detectors.