KagglePlanetPytorch

This repository contains the basic code of our 9th place submission.

For questions refere to https://www.kaggle.com/c/planet-understanding-the-amazon-from-space/discussion/36887 or create an issue!

Requirements

Basic requirements are

• Scitkit-Learn
• Scikit-Image
• Numpy, Scipy
• Torchsample
• Pytorch
• XGBoost

This list may not be exhaustive!

Training a network

Just run the nn_finetune-files.

Create predictions for a network

Choose the network in predict.py and run it. Predictions are then saved to /predictions.

Calculate thresholds

This step is only necessary because of the current implementation. Run save_thresholds.py for your model. The saved thresholds we be used in the next step to compare XGBoost to averaging.

Make a submission from a single 5-fold model

Specify the network in model_tta_hyperopt.py and run it. This will run hyper parameter optmization for XGBoost. The approach chosen in this file is probably not good at all, since this was the first time I used XGBoost and only had a week to the competition deadline. Please tell me if you can do better. Also if you can make the same basic approach work for model ensembling, tell me! :) Submission are saved to /submissions

Make a weighted submissions from different submission files

Just specify your submissions and weights in submit_ensemble.py