# 33rd Place Solution

First of all thanks to competition organizer and Kaggle to host such interesting competition. For the first time I learned about audio features and how ML is applied on them. Quite a learning.
Whatever improvement I got is during last 4 days.
Previous to that I struggled to match the best public kernel.
Anyways the score is of single fold Resnest50 as shared in public kernel.
Training process: key points
Extracted nocall (duration for which no labels are present) from example_test_audio and BIRDCLEF 2020 data and used as noise as well as extra nocall class.
As ResNet50 expect images with 3 channel and MelSpectrogram gives single channel I stacked three different Melspectrogram with window size of 1024,1536 ans 2048. Not sure is it a good technique as I have not evaluated the same model and training process with same channel duplicated.
Inference:
Same as available in public kernel with threshold of 0.5
Lots of areas for improvement
One idea was to select sample from competition training data based on available metadata like rating, duration, background birds.
Training more models and ensemble.
At the end local CV (CV of random clips from validation data) matches the private leaderboard.
Also  I think I am bit lucky as I see quite a lot pvt leaderboard shakeup.