{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/explaining-deep-convolutional-neural-networks","title":"Explaining Deep Convolutional Neural Networks on Music Classification","arxiv_id":"1607.02444","date":"2016-07-08","proceeding":null,"authors":["Keunwoo Choi","George Fazekas","Mark Sandler"],"abstract":"Deep convolutional neural networks (CNNs) have been actively adopted in the\nfield of music information retrieval, e.g. genre classification, mood\ndetection, and chord recognition. However, the process of learning and\nprediction is little understood, particularly when it is applied to\nspectrograms. We introduce auralisation of a CNN to understand its underlying\nmechanism, which is based on a deconvolution procedure introduced in [2].\nAuralisation of a CNN is converting the learned convolutional features that are\nobtained from deconvolution into audio signals. In the experiments and\ndiscussions, we explain trained features of a 5-layer CNN based on the\ndeconvolved spectrograms and auralised signals. The pairwise correlations per\nlayers with varying different musical attributes are also investigated to\nunderstand the evolution of the learnt features. It is shown that in the deep\nlayers, the features are learnt to capture textures, the patterns of continuous\ndistributions, rather than shapes of lines.","url_abs":"http://arxiv.org/abs/1607.02444v1","url_pdf":"http://arxiv.org/pdf/1607.02444v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"explaining-deep-convolutional-neural-networks","repo_url":"https://github.com/keunwoochoi/Auralisation","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok"}}],"tasks":[{"task_slug":"chord-recognition","task_name":"Chord Recognition"},{"task_slug":"classification-1","task_name":"Classification"},{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"genre-classification","task_name":"Genre classification"},{"task_slug":"information-retrieval","task_name":"Information Retrieval"},{"task_slug":"music-classification","task_name":"Music Classification"},{"task_slug":"music-information-retrieval","task_name":"Music Information Retrieval"},{"task_slug":"retrieval","task_name":"Retrieval"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}