My modified version of YoloV5 training, cross-validation and inference with Pseudo Labelling pytorch pipelines used in GWD Kaggle Competition
-
Updated
Aug 28, 2020 - Jupyter Notebook
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
Label-free inference of latent intra-annual multi-cropping sequences from multi-sensor SITS: Fourier-basis gap reconstruction, batched bounded LM double-logistic inversion, α-threshold phenometrics, rule-induced pseudo-labels, spatially-blocked RF/TempCNN, MK–Theil–Sen operators, FAO-56 ET accounting
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.
To associate your repository with the pseudo-labelling topic, visit your repo's landing page and select "manage topics."