This repo is for Dacon competition. After competition, all submitted codes will be opened.
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.
$ pip install -r requirements.txt
To understand basic pipeline, check toy.py
.
$ python toy.py
It contains simple CNN as base model. You can customize it simply.
Following without any changes, prediction results about test data exported in ./results
.
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.