Skip to content
6th place solution
Branch: master
Clone or download
Latest commit c0b143d Apr 30, 2017
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
LICENSE Initial commit Mar 9, 2017
README.md Update README.md Apr 30, 2017
b3_data_iter.py add sources Mar 15, 2017
b3_gen_prediction.py add all files Apr 30, 2017
b3_train.py
blend_raw.py add all files Apr 30, 2017
multi_predict.py add all files Apr 30, 2017
nn_predict.py add all files Apr 30, 2017
predict_utils.py add all files Apr 30, 2017
raw_to_mask.py add all files Apr 30, 2017
simplify_sub.py add all files Apr 30, 2017
utils.py add utils Apr 30, 2017
view_pred_poly.py add all files Apr 30, 2017

README.md

kaggle-dstl-satellite-imagery-feature-detection

Scripts for the my 6th place solution with MXNet.
Key features:

  • Unet-like network architecture with multiple input branches(A, M, P channels)
  • Adaptive crop sampler, based on the performance of the network on different classes, updated each epoch.

Brief description:
b3_data_iter - data generation, augmentation
utils - wkt polygon rasterization, raster polygonization utils
b3_train - network symbol definition, training loop

You can’t perform that action at this time.