Freesound Audio Tagging
https://www.kaggle.com/c/freesound-audio-tagging
1d_final_launch:
Training 10 1d models, saving predictions
2d_final_launch:
Training 10 2d models, saving predictions
ensembling_submit - calculates the geometric mean of the predictions made on the previous two
steps.
Network architectures templates for experiments: archs_2.py.
Template to conduct experiments on choosing network architecture - pipeline_general.ipynb
Evaluation
MAP@3
Result: 0.91