Beitadoge / kaggle_salt_bes_phalanx

Kaggle Salt b.e.s. & phalanx

Kaggle TGS Salt Identification Challenge. b.e.s. & phalanx 1st Place Solution

To read the detailed solution, please, refer to the Kaggle post

ENVIRONMENT

The solution is available as a Docker container. The following dependecies should be installed:

• Python 3.5.2
• CUDA 9.0
• cuddn 7
• nvidia drivers v.384
• Docker
• nvidia-docker

DATA SETUP

Download and unzip competition data into data/ directory. One could specify local path to the new test images in SETTINGS.json file (NEW_TEST_IMAGES_DATA field). The competition test data is used by default.

DOCKER SETUP

To build and start a docker container run:

cd docker 
./build.sh
./run.sh

MODEL BUILD

1. train models

   a) expect this to run for about 16 days on a single GTX1080Ti

   b) trains all models from scratch

2. make prediction

   a) expect this to run for 3.5 hours for 18,000 test images on a single GTX1080Ti

   b) uses saved model weights

Commands to run each build are presented below:

1. train models (creates model weights in bes/weights and phalanx/weights)

/.train.sh

2. ordinary prediction (creates predictions/test_prediction.csv)

./predict.sh

ADDITIONAL NOTES

1. Model weights are saved in bes/weights and phalanx/weights for b.e.s. and phalanx models respectively

2. Individual model predictions before ensembling are stored in bes/predictions (lots of .png images) and phalanx/predictions (.npy files)

3. Scripts to generate initial folds and jigsaw mosaics are located in bes/datasets: generate_folds.py and Generate_Mosaic.R