This is repository of the 4th place solution of kaggleFreesound Audio Tagging 2019 competition.
The discription of this solution is available at
http://dcase.community/challenge2019/task-audio-tagging-results#Akiyama2019
https://www.kaggle.com/c/freesound-audio-tagging-2019/discussion/96440

• Python 3.6.6
• CUDA 10.0
• numpy (1.16.4)
• pandas (0.23.4)
• matplotlib (3.1.0)
• Pytorch (1.1.0)
• librosa (0.6.3)
• sci-kit learn (0.21.2)
• scipy (1.2.1)
• pretrainedmodels (0.7.4)

Download the dataset and place them in input/.
Unzip zip files and place them to train_curated/, train_noisy/, test/.
In case you use pretrained weights, download the weights, unzip zipped weights and place them to models/resnet_model1/, models/resnet_model2/ and so on.

Run src/preprocess.py.
Run src/train_model1.py.
Run src/get_pseudo_label.py.
Run src/train_model2.py .
Run src/train_model3.py .
Run src/train_model4_0.py.
Run src/train_model4.py.
Run src/train_model5.py.
Run src/train_model6_0.py.
Run src/train_model6.py.

Run src/make_final_submission1.py. The submission file output/submission1.csv will be generted.
Run src/make_final_submission2.py. . The submission file output/submission2.csv will be generted.