{ "id": 103923, "title": "And the Judges' Award goes to ...", "url": "/competitions/freesound-audio-tagging-2019/discussion/103923", "author_name": "", "post_date": "2019-08-12T23:54:01.825691Z", "votes": 9, "comment_count": 2, "views": 0, "content": "

Dear all,

\n\n

After reviewing the technical reports submitted to the DCASE2019 Challenge Task2, the Judges’ Award goes to:

\n\n

Eric Bouteillon ( @ebouteillon ) and his submission

\n\n

SpecMix: A Simple Data Augmentation and Warm-up Pipeline to Leverage Clean and Noisy Set for Efficient Audio Tagging

\n\n

A simple yet effective and interesting methodology that considers the aspects of the FSDKaggle2019 dataset and that has attained competitive results with limited computational requirements. We also note his effort in documenting and sharing his code for reproducibility. Congratulations Eric, on behalf of the organizers!

\n\n

In addition, we’d like to mention the following two teams that have also proposed very interesting approaches.\n- CP-JKU ( @khalvic ) with their submission CP-JKU: Submissions to DCASE’19: Acoustic Scene Classification and Audio Tagging with Receptive-Field-Regularized CNNs.\n- 3x6min ( @jihangz @lftuwujie ) with their submission DCASE 2019 TASK 2: Semi-Supervised Networks with Heavy Data Augmentions to Battle Against Label Noise in Audio Tagging Task.

\n\n

With this, the Freesound Audio Tagging Challenge 2019 is concluded. On behalf of the organizers, thank you all for your passionate participation. We hope that you enjoyed this scientific evaluation as much as we did, and that it allowed you to face new challenges in the recognition of sound events. You can check out the DCASE2019 Challenge Task2 results page for the list of technical reports, writeups and code repositories released so far in this Challenge.

\n\n

-- The organizers\n - Eduardo Fonseca, MTG-UPF, Barcelona https://twitter.com/edfonseca_

\n\n", "messages": [ { "id": "597939", "postDate": "08/12/2019 23:54:01", "content": "

Dear all,

\n\n

After reviewing the technical reports submitted to the DCASE2019 Challenge Task2, the Judges’ Award goes to:

\n\n

Eric Bouteillon ( @ebouteillon ) and his submission

\n\n

SpecMix: A Simple Data Augmentation and Warm-up Pipeline to Leverage Clean and Noisy Set for Efficient Audio Tagging

\n\n

A simple yet effective and interesting methodology that considers the aspects of the FSDKaggle2019 dataset and that has attained competitive results with limited computational requirements. We also note his effort in documenting and sharing his code for reproducibility. Congratulations Eric, on behalf of the organizers!

\n\n

In addition, we’d like to mention the following two teams that have also proposed very interesting approaches.\n- CP-JKU ( @khalvic ) with their submission CP-JKU: Submissions to DCASE’19: Acoustic Scene Classification and Audio Tagging with Receptive-Field-Regularized CNNs.\n- 3x6min ( @jihangz @lftuwujie ) with their submission DCASE 2019 TASK 2: Semi-Supervised Networks with Heavy Data Augmentions to Battle Against Label Noise in Audio Tagging Task.

\n\n

With this, the Freesound Audio Tagging Challenge 2019 is concluded. On behalf of the organizers, thank you all for your passionate participation. We hope that you enjoyed this scientific evaluation as much as we did, and that it allowed you to face new challenges in the recognition of sound events. You can check out the DCASE2019 Challenge Task2 results page for the list of technical reports, writeups and code repositories released so far in this Challenge.

\n\n

-- The organizers\n - Eduardo Fonseca, MTG-UPF, Barcelona https://twitter.com/edfonseca_

\n\n", "rawMarkdown": "Dear all, \n\nAfter reviewing the technical reports submitted to the DCASE2019 Challenge Task2, the Judges’ Award goes to:\n\n **Eric Bouteillon** ( @ebouteillon ) and his submission \n\n**SpecMix: A Simple Data Augmentation and Warm-up Pipeline to Leverage Clean and Noisy Set for Efficient Audio Tagging**\n\nA simple yet effective and interesting methodology that considers the aspects of the FSDKaggle2019 dataset and that has attained competitive results with limited computational requirements. We also note his effort in documenting and sharing his code for reproducibility. Congratulations Eric, on behalf of the organizers! \n\nIn addition, we’d like to mention the following two teams that have also proposed very interesting approaches.\n- **CP-JKU** ( @khalvic ) with their submission CP-JKU: Submissions to DCASE’19: Acoustic Scene Classification and Audio Tagging with Receptive-Field-Regularized CNNs.\n- **3x6min** ( @jihangz @lftuwujie ) with their submission DCASE 2019 TASK 2: Semi-Supervised Networks with Heavy Data Augmentions to Battle Against Label Noise in Audio Tagging Task.\n\nWith this, the **Freesound Audio Tagging Challenge 2019** is concluded. On behalf of the organizers, thank you all for your passionate participation. We hope that you enjoyed this scientific evaluation as much as we did, and that it allowed you to face new challenges in the recognition of sound events. You can check out the DCASE2019 Challenge Task2 results page for the list of technical reports, writeups and code repositories released so far in this Challenge.\n\n-- The organizers\n - Eduardo Fonseca, MTG-UPF, Barcelona https://twitter.com/edfonseca_\n\n - Manoj Plakal, Google's Sound Understanding, New York\n\n - Frederic Font, MTG-UPF, Barcelona\n\n - Dan Ellis, Google's Sound Understanding, New York", "votes": null }, { "id": "600243", "postDate": "08/15/2019 20:39:08", "content": "

Thanks to all the organisers for preparing this competition and reviewing all the technical reports submitted. It was a surprise for me to receive this award as other reports were excellent. I discovered very interesting things in them, therefore I highly recommend to visit the dcase website to get the submitted technical reports from other competitors. \nAs a side note, I want to say that I consider sharing knowledge important, that is why I shared in previous competition what I did and in this competition too. I  will continue to contribute back to the Kaggle community in next competitions I will participate in.

", "rawMarkdown": "Thanks to all the organisers for preparing this competition and reviewing all the technical reports submitted. It was a surprise for me to receive this award as other reports were excellent. I discovered very interesting things in them, therefore I highly recommend to visit the dcase website to get the submitted technical reports from other competitors. \nAs a side note, I want to say that I consider sharing knowledge important, that is why I shared in previous competition what I did and in this competition too. I  will continue to contribute back to the Kaggle community in next competitions I will participate in.", "votes": null }, { "id": "600467", "postDate": "08/16/2019 06:25:17", "content": "

Congratulation Eric!!

", "rawMarkdown": "Congratulation Eric!!", "votes": null } ], "comments": [ { "id": 600243, "author_name": "ebouteillon", "author_url": "", "post_date": "08/15/2019 20:39:08", "content": "

Thanks to all the organisers for preparing this competition and reviewing all the technical reports submitted. It was a surprise for me to receive this award as other reports were excellent. I discovered very interesting things in them, therefore I highly recommend to visit the dcase website to get the submitted technical reports from other competitors. \nAs a side note, I want to say that I consider sharing knowledge important, that is why I shared in previous competition what I did and in this competition too. I  will continue to contribute back to the Kaggle community in next competitions I will participate in.

", "votes": null, "replies": [ { "id": 600467, "author_name": "ratthachat", "author_url": "", "post_date": "08/16/2019 06:25:17", "content": "

Congratulation Eric!!

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "597939": "Dear all, \n\nAfter reviewing the technical reports submitted to the DCASE2019 Challenge Task2, the Judges’ Award goes to:\n\n **Eric Bouteillon** ( @ebouteillon ) and his submission \n\n**SpecMix: A Simple Data Augmentation and Warm-up Pipeline to Leverage Clean and Noisy Set for Efficient Audio Tagging**\n\nA simple yet effective and interesting methodology that considers the aspects of the FSDKaggle2019 dataset and that has attained competitive results with limited computational requirements. We also note his effort in documenting and sharing his code for reproducibility. Congratulations Eric, on behalf of the organizers! \n\nIn addition, we’d like to mention the following two teams that have also proposed very interesting approaches.\n- **CP-JKU** ( @khalvic ) with their submission CP-JKU: Submissions to DCASE’19: Acoustic Scene Classification and Audio Tagging with Receptive-Field-Regularized CNNs.\n- **3x6min** ( @jihangz @lftuwujie ) with their submission DCASE 2019 TASK 2: Semi-Supervised Networks with Heavy Data Augmentions to Battle Against Label Noise in Audio Tagging Task.\n\nWith this, the **Freesound Audio Tagging Challenge 2019** is concluded. On behalf of the organizers, thank you all for your passionate participation. We hope that you enjoyed this scientific evaluation as much as we did, and that it allowed you to face new challenges in the recognition of sound events. You can check out the DCASE2019 Challenge Task2 results page for the list of technical reports, writeups and code repositories released so far in this Challenge.\n\n-- The organizers\n - Eduardo Fonseca, MTG-UPF, Barcelona https://twitter.com/edfonseca_\n\n - Manoj Plakal, Google's Sound Understanding, New York\n\n - Frederic Font, MTG-UPF, Barcelona\n\n - Dan Ellis, Google's Sound Understanding, New York", "600243": "Thanks to all the organisers for preparing this competition and reviewing all the technical reports submitted. It was a surprise for me to receive this award as other reports were excellent. I discovered very interesting things in them, therefore I highly recommend to visit the dcase website to get the submitted technical reports from other competitors. \nAs a side note, I want to say that I consider sharing knowledge important, that is why I shared in previous competition what I did and in this competition too. I  will continue to contribute back to the Kaggle community in next competitions I will participate in.", "600467": "Congratulation Eric!!" }, "source": "meta" }