Kaggle에서 찾은 Pseudo labeling 전략

[lsh3163]

1. 앙상블한 박스들 중 threshold (0.4, 0.5, 0.05) 이상인 박스들을 새로운 데이터에 추가합니다!
   • https://www.kaggle.com/kaushal2896/efficientdet-pseudo-labeling-wbf-tta
   • https://www.kaggle.com/nvnnghia/fasterrcnn-pseudo-labeling
   • https://www.kaggle.com/nvnnghia/yolov5-pseudo-labeling/notebook#Make-pseudo-labels-for-Yolov5