Introduction

This repo is for Dacon competition. After competition, all submitted codes will be opened.

Prerequisite

Data Preparation

First of all, you should download data from Dacon Competition site. You will get 4 files which are train.csv, test.csv, sample_submission.csv, labels_mapping.csv and move these files into {project_root_dir}/resource/data/. At first, there's no resource directory, you may create directory using like mkdir command or any other method preferred.

Package Installation

$ pip install -r requirements.txt

Execution

To understand basic pipeline, check toy.py.

$ python toy.py

It contains simple CNN as base model. You can customize it simply.

Results

Following without any changes, prediction results about test data exported in ./results.

Competition Records

Official Leaderboard

Our team name was SRiracha, we ranked on 21th place on private dataset. We finally use ensemble model with previous submitted model. Our model scored(MacroF1) 0.78056 on private data.

Contributor

• Eunsik Lee(@emphasis10)