GithubHelp home page GithubHelp logo

yuv4r4j / kaggle_salt_bes_phalanx Goto Github PK

View Code? Open in Web Editor NEW

This project forked from ybabakhin/kaggle_salt_bes_phalanx

0.0 1.0 0.0 6.84 MB

Kaggle Salt b.e.s. & phalanx

Python 91.09% R 3.02% Shell 5.63% Dockerfile 0.26%

kaggle_salt_bes_phalanx's Introduction

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

ArXiv preprint with the detailed description of the solution:
Semi-Supervised Segmentation of Salt Bodies in Seismic Images using an Ensemble of Convolutional Neural Networks
Yauhen Babakhin, Artsiom Sanakoyeu, Hirotoshi Kitamura
https://arxiv.org/abs/1904.04445

Kaggle post about the solution: link.

ENVIRONMENT

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

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.

WEIGHTS SETUP

To get the weights from the final stage models download them from google drive and unzip into corresponding bes/weights/ and phalanx/weights directories.

DOCKER SETUP

To build and start a docker container run:

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

MODEL BUILD

  1. train models from scratch

    a) trains all models from scratch

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

  2. make prediction

    a) uses weights from the final stage models to make predictions

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

Commands to run each build are presented below:

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

./train.sh

2. make 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

CITATION

If you find this code useful, please cite the preprint:

@journal{tgsSaltBodiesSegmentation2019,
  title={Semi-Supervised Segmentation of Salt Bodies in Seismic Images using an Ensemble of Convolutional Neural Networks},
  author={ Babakhin, Yauhen, and Sanakoyeu, Artsiom, and Kitamura, Hirotoshi},
  journal={arXiv preprint arXiv:1904.04445},
  year={2019}
}

kaggle_salt_bes_phalanx's People

Contributors

ybabakhin avatar asanakoy avatar

Watchers

James Cloos avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.