# 79th Place Solution. 🥈

Dear Kagglers.
Thank you to the competition organizers hosting this interesting competition.
Thank you to everyone involved in this competition. I learned a lot from public notebooks and discussions.
My Final Result
Score
Public: 0.751 → Private 0.786
LeaderBoard
Public: 52nd → Private 79th 🥈
Summary of Solution
![](assets/012e92864856418e90676027.png)
common settings
criterion: focal loss
1DUnet
based on these
[@werus23](https://www.kaggle.com/werus23)
's and
[@danielphalen](https://www.kaggle.com/danielphalen)
's notebooks, I modified loss criterion, dataset, features, architecture, etc.
[https://www.kaggle.com/code/werus23/sleep-critical-point-train](https://www.kaggle.com/code/werus23/sleep-critical-point-train)
[https://www.kaggle.com/code/werus23/sleep-critical-point-infer](https://www.kaggle.com/code/werus23/sleep-critical-point-infer)
[https://www.kaggle.com/code/danielphalen/cmss-grunet-train](https://www.kaggle.com/code/danielphalen/cmss-grunet-train)
[https://www.kaggle.com/code/danielphalen/cmss-grunet-infer](https://www.kaggle.com/code/danielphalen/cmss-grunet-infer)
architecture
![](assets/65014b55c2e1a2e889f1ec06.png)
Spec2DCNN
based on
[@tubotubo](https://www.kaggle.com/tubotubo)
's public repository, I added focal loss criterion.
[https://github.com/tubo213/kaggle-child-mind-institute-detect-sleep-states](https://github.com/tubo213/kaggle-child-mind-institute-detect-sleep-states)
[https://www.kaggle.com/code/tubotubo/cmi-submit](https://www.kaggle.com/code/tubotubo/cmi-submit)
Source (inference notebook)
[https://www.kaggle.com/code/hutch1221/cmi-ensemble](https://www.kaggle.com/code/hutch1221/cmi-ensemble)
Finally
I gained significant insights into event detection techniques from the public notebooks and discussions. In particular, the public codes shared by
[@werus23](https://www.kaggle.com/werus23)
,
[@danielphalen](https://www.kaggle.com/danielphalen)
, and
[@tubotubo](https://www.kaggle.com/tubotubo)
were outstanding and contributed greatly to my learning.
Advice is always welcome!
Thank you for your attention.