Since, spectrogram helps to get better scores, its good to explore the top solutions of previous competitions which applied similar to spectrogram approach ( mainly bird competitions as i participated, share other competitions where spectorgram used too )
\nUniversity of Liverpool - Ion Switching
\nWill update more as i go through more solutions. Share if you come across good top solutions.
", "messages": [ { "id": "2602069", "postDate": "01/14/2024 21:25:18", "content": "Since, spectrogram helps to get better scores, its good to explore the top solutions of previous competitions which applied similar to spectrogram approach ( mainly bird competitions as i participated, share other competitions where spectorgram used too )
\nUniversity of Liverpool - Ion Switching
\nWill update more as i go through more solutions. Share if you come across good top solutions.
", "rawMarkdown": "Since, spectrogram helps to get better scores, its good to explore the top solutions of previous competitions which applied similar to spectrogram approach ( mainly bird competitions as i participated, share other competitions where spectorgram used too )\n\n# Spectrogram Based\n[BirdCLEF 2023](https://github.com/moritake04/birdclef-2023)\n- [Discussion - 20th place solution: SED + CNN ensemble using onnx](https://www.kaggle.com/competitions/birdclef-2023/overview) **[Code](https://github.com/moritake04/birdclef-2023)**\n\n[BirdCLEF 2022](https://www.kaggle.com/competitions/birdclef-2022/overview)\n- [Discussion - 1st place solution models (it’s not all BirdNet)](https://www.kaggle.com/competitions/birdclef-2022/discussion/327047) **[Code](https://github.com/Selimonder/birdclef-2022)**\n\n# EGG (Wav) Based\n[University of Liverpool - Ion Switching](https://www.kaggle.com/c/liverpool-ion-switching/overview)\n- [Discussion - 7th Place - Part 2 - Magical WaveNet ](https://www.kaggle.com/c/liverpool-ion-switching/discussion/154253)\n- [Discussion - 5th Place Solution with Code](https://www.kaggle.com/competitions/liverpool-ion-switching/discussion/154220) **[Code](https://www.kaggle.com/meminozturk/into-the-wild-wavenet/)** **go through discussion**\n\nWill update more as i go through more solutions. Share if you come across good top solutions.", "votes": null }, { "id": "2606912", "postDate": "01/17/2024 23:39:55", "content": "Updated Wavnet based previous solutions. @cdeotte is possible to see the code for Magical WaveNet solution.
", "rawMarkdown": "Updated Wavnet based previous solutions. @cdeotte is possible to see the code for **Magical WaveNet solution**.", "votes": null }, { "id": "2606955", "postDate": "01/18/2024 01:12:49", "content": "Ion Comp here is also a good example of WaveNet solutions. The winning solutions, stacked WaveNet over GBT. So for example, we can infer GBT on spectrograms. Then input the output of GBT plus EEG Waveform into WaveNet
", "rawMarkdown": "Ion Comp [here][1] is also a good example of WaveNet solutions. The winning solutions, stacked WaveNet over GBT. So for example, we can infer GBT on spectrograms. Then input the output of GBT plus EEG Waveform into WaveNet\n\n[1]: https://www.kaggle.com/competitions/liverpool-ion-switching", "votes": null }, { "id": "2613275", "postDate": "01/22/2024 01:04:11", "content": "Isn't WaveNet designed to produce audio, in this case signals. I have briefly checked out ion comp and the prediction seems also to be in serial, a continuous prediction of opening and closing pump. So the form of prediction is quite different here. ccorrect me if I am wrong, @cdeotte
", "rawMarkdown": "Isn't WaveNet designed to produce audio, in this case signals. I have briefly checked out ion comp and the prediction seems also to be in serial, a continuous prediction of opening and closing pump. So the form of prediction is quite different here. ccorrect me if I am wrong, @cdeotte", "votes": null } ], "comments": [ { "id": 2606912, "author_name": "seshurajup", "author_url": "", "post_date": "01/17/2024 23:39:55", "content": "Updated Wavnet based previous solutions. @cdeotte is possible to see the code for Magical WaveNet solution.
", "votes": null, "replies": [ { "id": 2606955, "author_name": "cdeotte", "author_url": "", "post_date": "01/18/2024 01:12:49", "content": "Ion Comp here is also a good example of WaveNet solutions. The winning solutions, stacked WaveNet over GBT. So for example, we can infer GBT on spectrograms. Then input the output of GBT plus EEG Waveform into WaveNet
", "votes": null, "replies": [ { "id": 2613275, "author_name": "renyiwei", "author_url": "", "post_date": "01/22/2024 01:04:11", "content": "Isn't WaveNet designed to produce audio, in this case signals. I have briefly checked out ion comp and the prediction seems also to be in serial, a continuous prediction of opening and closing pump. So the form of prediction is quite different here. ccorrect me if I am wrong, @cdeotte
", "votes": null, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "2602069": "Since, spectrogram helps to get better scores, its good to explore the top solutions of previous competitions which applied similar to spectrogram approach ( mainly bird competitions as i participated, share other competitions where spectorgram used too )\n\n# Spectrogram Based\n[BirdCLEF 2023](https://github.com/moritake04/birdclef-2023)\n- [Discussion - 20th place solution: SED + CNN ensemble using onnx](https://www.kaggle.com/competitions/birdclef-2023/overview) **[Code](https://github.com/moritake04/birdclef-2023)**\n\n[BirdCLEF 2022](https://www.kaggle.com/competitions/birdclef-2022/overview)\n- [Discussion - 1st place solution models (it’s not all BirdNet)](https://www.kaggle.com/competitions/birdclef-2022/discussion/327047) **[Code](https://github.com/Selimonder/birdclef-2022)**\n\n# EGG (Wav) Based\n[University of Liverpool - Ion Switching](https://www.kaggle.com/c/liverpool-ion-switching/overview)\n- [Discussion - 7th Place - Part 2 - Magical WaveNet ](https://www.kaggle.com/c/liverpool-ion-switching/discussion/154253)\n- [Discussion - 5th Place Solution with Code](https://www.kaggle.com/competitions/liverpool-ion-switching/discussion/154220) **[Code](https://www.kaggle.com/meminozturk/into-the-wild-wavenet/)** **go through discussion**\n\nWill update more as i go through more solutions. Share if you come across good top solutions.", "2606912": "Updated Wavnet based previous solutions. @cdeotte is possible to see the code for **Magical WaveNet solution**.", "2606955": "Ion Comp [here][1] is also a good example of WaveNet solutions. The winning solutions, stacked WaveNet over GBT. So for example, we can infer GBT on spectrograms. Then input the output of GBT plus EEG Waveform into WaveNet\n\n[1]: https://www.kaggle.com/competitions/liverpool-ion-switching", "2613275": "Isn't WaveNet designed to produce audio, in this case signals. I have briefly checked out ion comp and the prediction seems also to be in serial, a continuous prediction of opening and closing pump. So the form of prediction is quite different here. ccorrect me if I am wrong, @cdeotte" }, "source": "meta" }