# 42 solution with code

thanks to the hosts, kaggle
thanks to
[@kkiller](https://www.kaggle.com/kkiller)
and his awosome code
[https://www.kaggle.com/kneroma/clean-fast-simple-bird-identifier-training-colab](https://www.kaggle.com/kneroma/clean-fast-simple-bird-identifier-training-colab)
3 step
train different model (resnet34, resnet50,resnest50,resnext50_32x4d)
find best ensemble model, depend on train_soundscapes
post preprocess, use season and position to filter bird
train
I use kkiller public datasets, and I try to add some augument, but not increase much, less 0.01
I add second label, with prob 0.3, have some improvement, I  study a lot last year's solution
[https://www.kaggle.com/c/birdsong-recognition/discussion/183208](https://www.kaggle.com/c/birdsong-recognition/discussion/183208)
[https://www.kaggle.com/c/birdsong-recognition/discussion/183269](https://www.kaggle.com/c/birdsong-recognition/discussion/183269)
[https://www.kaggle.com/c/birdsong-recognition/discussion/183199](https://www.kaggle.com/c/birdsong-recognition/discussion/183199)
[https://www.kaggle.com/c/birdsong-recognition/discussion/183204](https://www.kaggle.com/c/birdsong-recognition/discussion/183204)
I also tried effientnet b0~b5, not work, I don't known why, wished  players can give some advice
ensemble
I trained diffent model, and use train_soundscapes f1 to check best epoch
I choose 2 sulotion, one for best local f1, and one add some random, use model not sort by local f1
post preprocess
I use uber h3 calculate Latitude and Longitude to get segment id, and use train_metadata.csv to get bird belong which segment, this give 0.01 improve
finally
thanks all, I am NLPer, not very familier with sound engineer, but kindly public code helped me a lot
my inference code:
[https://www.kaggle.com/jt120lz/clean-fast-simple-bird-identifier-inference](https://www.kaggle.com/jt120lz/clean-fast-simple-bird-identifier-inference)