@techreport{Arnault2020,
    Author = "Arnault, Augustin and Riche, Nicolas",
    title = "{CRNNs} for Urban Sound Tagging with Spatiotemporal Context",
    institution = "DCASE2020 Challenge",
    year = "2020",
    month = "October",
    abstract = "This paper describes CRNNs we used to participate in Task 5 of the DCASE 2020 challenge. This task focuses on hierarchical multilabel urban sound tagging with spatiotemporal context. The code is available to our GitHub repository at https://github.com/multitelai/urban-sound-tagging."
}

@techreport{Bai2020,
    Author = "Bai, Jisheng and Chen, Chen and Chen, Jianfeng and Wang, Mou and Zhang, Xiaolei and Yan, Qingli",
    title = "Data Augmentation Based System for Urban Sound Tagging",
    institution = "DCASE2020 Challenge",
    year = "2020",
    month = "October",
    abstract = "In this report, we introduce our system for Task5 of Dcase 2020 challenges (Urban Sound Tagging with Spatiotemporal Context). We presents a fusion system for Task5 based on different features and data augmentation. The task focuses on predicting whether each of 23 sources of noise pollution is present or absent in a 10- second scene with original recordings and addtional spatiotemporal context [1]. There are two levels of taxonomies to train a model. To explore features in detecting various sources of urban sound, we extracted four different features from original recordings. We applied a 9-layer convolutional neural network(CNN) as our primary classifier. To prevent the imbalance between classes, we applied data augmentation method. The experiment results show that our system performed better than baseline on validation data."
}

@techreport{Cartwright2020,
    Author = "Cartwright, Mark and Cramer, Jason and Mendez Mendez, Ana Elisa and Wang, Yu and Wu, Ho-Hsiang and Lostanlen, Vincent and Fuentes, Magdalena and Salamon, Justin and Bello, Juan P.",
    title = "An Urban Sound Tagging Dataset with Spatiotemporal Context",
    institution = "DCASE2020 Challenge",
    year = "2020",
    month = "October",
    abstract = "We present SONYC-UST v2, a dataset for urban sound tagging with spatiotemporal information. This dataset is aimed for the development and evaluation of machine listening systems for realworld urban noise monitoring. While datasets of urban recordings are available, this datasets provides the opportunity to investigate how spatiotemporal metadata can aid in the prediction of urban sound tags. It consists of 18510 audio recordings from the 'Sounds of New York City' (SONYC) acoustic sensor network including relevant metadata about when and where the data were recorded at the hour and block level. The dataset has annotations by volunteers from the Zooniverse citizen science platform, as well as a two-stage verification with our team. In this work we describe the data collection, metrics used to evaluate tagging systems, and the results of a simple baseline model that exploits temporal information."
}

@techreport{Diez2020,
    Author = "Diez, Itxasne and Gonzalez, Peio and Gonzalez, Ibon",
    title = "Urban Sound Classification Using Convolutional Neural Networks for {DCASE} 2020 Challenge",
    institution = "DCASE2020 Challenge",
    year = "2020",
    month = "October",
    abstract = "This technical report describes our system proposed for Task 5 - Urban Sound Tagging. The system has a core architecture based on Convolutional Neural Networks. This neural network uses log melspectrogram features as input and this input is processed by two CNN layers. The output of the convolutional stack is processed by several fully connected layers plus an output layer to produce the classification decision. Spatiotemporal context data is also available and we propose a multi-input architecture, with two input branches that are merged for the final processing. The spatiotemporal context information is processed by an additional neural network of 2 fully connected layers. Its output is merged with the output of the CNN stack and the resulting data is fed to the fully connected output block. In this report, we describe the proposed models in detail and compare them to the baseline approach using the provided development datasets. Finally, we present the results obtained with the validation split from the dataset."
}

@techreport{Iqbal2020,
    Author = "Iqbal, Turab and Cao, Yin and Plumbley, Mark D. and Wang, Wenwu",
    title = "Incorporating Auxiliary Data for Urban Sound Tagging",
    institution = "DCASE2020 Challenge",
    year = "2020",
    month = "October",
    abstract = "DCASE 2020 Task 5 presents a multi-label sound tagging problem for the detection of urban noises in acoustic scenes. The main theme is the use of auxiliary data to facilitate sound tagging. In this report, we provide a detailed description of our submission for Task 5 and present experimental results for the development set. Two different network choices are described: a pre-trained convolutional neural network (CNN) and a randomly-initialised gated CNN. To make use of the auxiliary information, we construct a feature vector based on the spatiotemporal metadata and use it in parallel with log-mel spectrogram features. Moreover, we address the presence of multiple annotations per recording by using a pseudo-labelling technique to estimate the true labels. Mean ensembling is also used with one of the proposed systems to combine several models."
}

@techreport{Kim2020,
    Author = "Kim, Jaehun",
    title = "Urban Sound Tagging Using Multi-Channel Audio Feature with Convolutional Neural Networks",
    institution = "DCASE2020 Challenge",
    year = "2020",
    month = "October",
    abstract = "This paper presents a multi-channel audio feature using imagenet model based on convolutional neural networks for DCASE 2020 Task5 Urban Sound Tagging (UST) with Spatio-temporal context (STC). We used the SONYC (Sounds of New York City) Urban Sound Tagging Dataset. It consists of audio clips and STC information. We proposed a multi-channel audio feature to use imagenet pre-trained model weight. multi-channel feature consists of raw and harmonic, percussive (HPSS) data’s Log-Mel-Spectrogram. Also, we used EfficientNet pre-trained model weight."
}

@techreport{Liu2020,
    Author = "Liu, Gang and Liu, Zhuangzhuang and Fang, Junyan and Hong, Xiaofeng",
    title = "Multisystem Fusion Model Based on Tag Relationship",
    institution = "DCASE2020 Challenge",
    year = "2020",
    month = "October",
    abstract = "Audio tagging aims to assign one or more labels to the audio clip. In this paper, we proposed our solutions applied to our submis-sion for DCASE2020 Task5 It focus on predicting whether each of the sources of noise pollution is present or absent in a 10-sec-ond scene [1]. And we should consider the spatiotemporal con-text(STC) in our work. We used VGG as our basic model and we regarded Multi-task learning as a method to train our models. We introduced the relationship between fine labels and coarse labels in our system. Finally, the coarse-grained and fine-grained taxon-omy results are obtained on the Micro Area under precision-recall curve (AUPRC), Micro F1 score and Macro Area under precision-recall curve (AUPRC)."
}