Just a couple of months ago I published an audio classification tutorial in R using xeno-canto data to classify among 50 bird species, which is very much in line with the goal of this competition. From that work I also created a Kaggle dataset. The underlying code is available from this GitHub repo. Notwithstanding the use of R, the use of a simple CNN model and the specification of the problem as multiclass classification - whereas the competition aims at multilabel - there might be nuggets of information that you can use.
\n\nI really would like to enter this competition but I do not have the time. I sincerely wish you all good luck!
", "messages": [ { "id": "895256", "postDate": "06/21/2020 07:57:26", "content": "Just a couple of months ago I published an audio classification tutorial in R using xeno-canto data to classify among 50 bird species, which is very much in line with the goal of this competition. From that work I also created a Kaggle dataset. The underlying code is available from this GitHub repo. Notwithstanding the use of R, the use of a simple CNN model and the specification of the problem as multiclass classification - whereas the competition aims at multilabel - there might be nuggets of information that you can use.
\n\nI really would like to enter this competition but I do not have the time. I sincerely wish you all good luck!
", "rawMarkdown": "Just a couple of months ago I published an [audio classification tutorial in R](https://poissonisfish.com/2020/04/05/audio-classification-in-r/) using xeno-canto data to classify among 50 bird species, which is very much in line with the goal of this competition. From that work I also created a [Kaggle dataset](https://www.kaggle.com/monogenea/birdsongs-from-europe). The underlying code is available from [this GitHub repo](https://github.com/monogenea/birdsong). Notwithstanding the use of R, the use of a simple CNN model and the specification of the problem as multiclass classification - whereas the competition aims at multilabel - there might be nuggets of information that you can use.\n\nI really would like to enter this competition but I do not have the time. I sincerely wish you all good luck!", "votes": null }, { "id": "895728", "postDate": "06/21/2020 15:09:14", "content": "Alles sind sehr gut. (tutorial/Dataset/GitHub)
", "rawMarkdown": "Alles sind sehr gut. (tutorial/Dataset/GitHub)", "votes": null }, { "id": "896031", "postDate": "06/21/2020 19:04:56", "content": "Danke Marília, Grüße
", "rawMarkdown": "Danke Marília, Grüße", "votes": null }, { "id": "897095", "postDate": "06/22/2020 16:03:00", "content": "Você já tentou usar TPU? No micro course Deep Learning há um módulo bonus para você experimentar. I highly recommend it. Try how fast is TPU. Since you're looking for being faster in your work. Maybe that can increase it even more. Much more.
", "rawMarkdown": "Você já tentou usar TPU? No micro course Deep Learning há um módulo bonus para você experimentar. I highly recommend it. Try how fast is TPU. Since you're looking for being faster in your work. Maybe that can increase it even more. Much more.", "votes": null }, { "id": "897167", "postDate": "06/22/2020 17:00:15", "content": "I heard of TPUs, I guess they require Python. I went through the Flower Classification tutorial to get a hold, still need to familiarise a bit more 🙂
", "rawMarkdown": "I heard of TPUs, I guess they require Python. I went through the Flower Classification tutorial to get a hold, still need to familiarise a bit more 🙂", "votes": null }, { "id": "897191", "postDate": "06/22/2020 17:29:37", "content": "I just made the Tutorial and exercise bonus lessons. It's worthy to read the discussions topics from Flower with TPU. The topics from Martin Görner, Dimitri Oliveira, Chris Deotte and Heng Cherkeng are very insightful.
", "rawMarkdown": "I just made the Tutorial and exercise bonus lessons. It's worthy to read the discussions topics from Flower with TPU. The topics from Martin Görner, Dimitri Oliveira, Chris Deotte and Heng Cherkeng are very insightful.", "votes": null } ], "comments": [ { "id": 895728, "author_name": "mpwolke", "author_url": "", "post_date": "06/21/2020 15:09:14", "content": "Alles sind sehr gut. (tutorial/Dataset/GitHub)
", "votes": null, "replies": [ { "id": 896031, "author_name": "monogenea", "author_url": "", "post_date": "06/21/2020 19:04:56", "content": "Danke Marília, Grüße
", "votes": null, "replies": [] } ] }, { "id": 897095, "author_name": "mpwolke", "author_url": "", "post_date": "06/22/2020 16:03:00", "content": "Você já tentou usar TPU? No micro course Deep Learning há um módulo bonus para você experimentar. I highly recommend it. Try how fast is TPU. Since you're looking for being faster in your work. Maybe that can increase it even more. Much more.
", "votes": null, "replies": [ { "id": 897167, "author_name": "monogenea", "author_url": "", "post_date": "06/22/2020 17:00:15", "content": "I heard of TPUs, I guess they require Python. I went through the Flower Classification tutorial to get a hold, still need to familiarise a bit more 🙂
", "votes": null, "replies": [] }, { "id": 897191, "author_name": "mpwolke", "author_url": "", "post_date": "06/22/2020 17:29:37", "content": "I just made the Tutorial and exercise bonus lessons. It's worthy to read the discussions topics from Flower with TPU. The topics from Martin Görner, Dimitri Oliveira, Chris Deotte and Heng Cherkeng are very insightful.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "895256": "Just a couple of months ago I published an [audio classification tutorial in R](https://poissonisfish.com/2020/04/05/audio-classification-in-r/) using xeno-canto data to classify among 50 bird species, which is very much in line with the goal of this competition. From that work I also created a [Kaggle dataset](https://www.kaggle.com/monogenea/birdsongs-from-europe). The underlying code is available from [this GitHub repo](https://github.com/monogenea/birdsong). Notwithstanding the use of R, the use of a simple CNN model and the specification of the problem as multiclass classification - whereas the competition aims at multilabel - there might be nuggets of information that you can use.\n\nI really would like to enter this competition but I do not have the time. I sincerely wish you all good luck!", "895728": "Alles sind sehr gut. (tutorial/Dataset/GitHub)", "896031": "Danke Marília, Grüße", "897095": "Você já tentou usar TPU? No micro course Deep Learning há um módulo bonus para você experimentar. I highly recommend it. Try how fast is TPU. Since you're looking for being faster in your work. Maybe that can increase it even more. Much more.", "897167": "I heard of TPUs, I guess they require Python. I went through the Flower Classification tutorial to get a hold, still need to familiarise a bit more 🙂", "897191": "I just made the Tutorial and exercise bonus lessons. It's worthy to read the discussions topics from Flower with TPU. The topics from Martin Görner, Dimitri Oliveira, Chris Deotte and Heng Cherkeng are very insightful." }, "source": "meta" }