pseudo-labelling

My modified version of YoloV5 training, cross-validation and inference with Pseudo Labelling pytorch pipelines used in GWD Kaggle Competition

My modified version of EfficientDet training, cross-validation and inference with Pseudo Labelling pytorch pipelines used in GWD Kaggle Competition

Ensemble of CatBoost + LightGBM (5-fold CV, class-weighted, pseudo-labeled with ~194K high-confidence test rows) using engineered color indices, redshift transforms, and leak-free KNN spatial-density features on cartesian sky coordinates; CV balanced accuracy 0.96833.