https://zindi.africa/competitions/inundata-mapping-floods-in-south-africa
Install flax, pandas, numpy
python code/run.py- augmentations, still heavy imbalance
- play with architecture
0.005731114 data aug, also lower batch size seems slightly better
0.006483697 on new bench with mods
0.022605287 on all 0s
0.183632111 baseline, simple XGBoost, prob still very imbalanced can play with hyperparams and better features no image yet...
Data: https://zindi.africa/competitions/inundata-mapping-floods-in-south-africa/data