In the surgical mask detection task, participants have to discriminate between utterances with and without surgical masks.
Participants have to train a model for surgical mask detection. This is a binary classification task in which an utterance (audio file) must be labeled as without mask (label 0) or with mask (label 1).
The training data is composed of 8000 audio files. The validation set is composed of 1000 audio files. The test is composed of 3000 audio files.
Kaggle competition: https://www.kaggle.com/c/ml-fmi-23-2020
My model accuracy is 0.69047 and I am on 6th place on the private leaderboard. https://www.kaggle.com/c/ml-fmi-23-2020/leaderboard
Read about the Approach on the wiki: Approach