{"file_date_updated":"2020-02-05T10:18:06Z","type":"conference","title":"Convolutional Recurrent Neural Network and Data Augmentation for Audio Tagging with Noisy Labels and Minimal Supervision","date_created":"2020-02-05T10:16:03Z","publication":"DCASE2019 Workshop, New York, USA","author":[{"last_name":"Ebbers","full_name":"Ebbers, Janek","id":"34851","first_name":"Janek"},{"full_name":"Haeb-Umbach, Reinhold","last_name":"Haeb-Umbach","id":"242","first_name":"Reinhold"}],"has_accepted_license":"1","status":"public","date_updated":"2023-11-22T08:30:12Z","year":"2019","quality_controlled":"1","_id":"15794","language":[{"iso":"eng"}],"ddc":["000"],"abstract":[{"lang":"eng","text":"In this paper we present our audio tagging system for the DCASE 2019 Challenge Task 2. We propose a model consisting of a convolutional front end using log-mel-energies as input features, a recurrent neural network sequence encoder and a fully connected classifier network outputting an activity probability for each of the 80 considered event classes. Due to the recurrent neural network, which encodes a whole sequence into a single vector, our model is able to process sequences of varying lengths. The model is trained with only little manually labeled training data and a larger amount of automatically labeled web data, which hence suffers from label noise. To efficiently train the model with the provided data we use various data augmentation to prevent overfitting and improve generalization. Our best submitted system achieves a label-weighted label-ranking average precision (lwlrap) of 75.5% on the private test set which is an absolute improvement of 21.7% over the baseline. This system scored the second place in the teams ranking of the DCASE 2019 Challenge Task 2 and the fifth place in the Kaggle competition “Freesound Audio Tagging 2019” with more than 400 participants. After the challenge ended we further improved performance to 76.5% lwlrap setting a new state-of-the-art on this dataset."}],"file":[{"file_size":184967,"date_updated":"2020-02-05T10:18:06Z","file_name":"DCASE_2019_WS_Ebbers_Paper.pdf","creator":"huesera","content_type":"application/pdf","relation":"main_file","file_id":"15795","access_level":"open_access","date_created":"2020-02-05T10:18:06Z"}],"department":[{"_id":"54"}],"project":[{"_id":"52","name":"Computing Resources Provided by the Paderborn Center for Parallel Computing"}],"user_id":"34851","oa":"1","citation":{"ieee":"J. Ebbers and R. Haeb-Umbach, “Convolutional Recurrent Neural Network and Data Augmentation for Audio Tagging with Noisy Labels and Minimal Supervision,” 2019.","apa":"Ebbers, J., &#38; Haeb-Umbach, R. (2019). Convolutional Recurrent Neural Network and Data Augmentation for Audio Tagging with Noisy Labels and Minimal Supervision. <i>DCASE2019 Workshop, New York, USA</i>.","mla":"Ebbers, Janek, and Reinhold Haeb-Umbach. “Convolutional Recurrent Neural Network and Data Augmentation for Audio Tagging with Noisy Labels and Minimal Supervision.” <i>DCASE2019 Workshop, New York, USA</i>, 2019.","bibtex":"@inproceedings{Ebbers_Haeb-Umbach_2019, title={Convolutional Recurrent Neural Network and Data Augmentation for Audio Tagging with Noisy Labels and Minimal Supervision}, booktitle={DCASE2019 Workshop, New York, USA}, author={Ebbers, Janek and Haeb-Umbach, Reinhold}, year={2019} }","ama":"Ebbers J, Haeb-Umbach R. Convolutional Recurrent Neural Network and Data Augmentation for Audio Tagging with Noisy Labels and Minimal Supervision. In: <i>DCASE2019 Workshop, New York, USA</i>. ; 2019.","chicago":"Ebbers, Janek, and Reinhold Haeb-Umbach. “Convolutional Recurrent Neural Network and Data Augmentation for Audio Tagging with Noisy Labels and Minimal Supervision.” In <i>DCASE2019 Workshop, New York, USA</i>, 2019.","short":"J. Ebbers, R. Haeb-Umbach, in: DCASE2019 Workshop, New York, USA, 2019."}}