{ "id": 264207, "title": "Similar Previous Competitions Top solutions", "url": "/competitions/g2net-gravitational-wave-detection/discussion/264207", "author_name": "", "post_date": "2021-08-11T11:23:41.533411600Z", "votes": 19, "comment_count": 1, "views": 0, "content": "

Previous top solutions of the similar Competitions

\n

BirdCLEF 2021 - Birdcall Identification

\n

1st place solution given by @startjapan,

\n\n

2nd place solution given by @philippsinger,

\n\n

3rd place solution given by @ludovick,

\n\n

4th place solution given by @tattaka,

\n\n

Freesound Audio Tagging 2019

\n

\"1\"
\n1st place solution given by @romul0212,

\n\n

2nd place solution given by @action,

\n\n

3rd place solution given by @ddanevskyi,

\n\n

4th place solution given by @osciiart,

\n\n

Bird audio detection

\n

1st place solution given by @umarzubair95,

\n\n

Cornell Birdcall Identification

\n

1st place solution given by @taggatle,

\n\n

2nd place solution given by @vlomme

\n\n

3rd place solution given by @theoviel,

\n", "messages": [ { "id": "1466240", "postDate": "08/11/2021 11:23:41", "content": "

Previous top solutions of the similar Competitions

\n

BirdCLEF 2021 - Birdcall Identification

\n

1st place solution given by @startjapan,

\n\n

2nd place solution given by @philippsinger,

\n\n

3rd place solution given by @ludovick,

\n\n

4th place solution given by @tattaka,

\n\n

Freesound Audio Tagging 2019

\n

\"1\"
\n1st place solution given by @romul0212,

\n\n

2nd place solution given by @action,

\n\n

3rd place solution given by @ddanevskyi,

\n\n

4th place solution given by @osciiart,

\n\n

Bird audio detection

\n

1st place solution given by @umarzubair95,

\n\n

Cornell Birdcall Identification

\n

1st place solution given by @taggatle,

\n\n

2nd place solution given by @vlomme

\n\n

3rd place solution given by @theoviel,

\n", "rawMarkdown": "`Previous top solutions of the similar Competitions`\n#[BirdCLEF 2021 - Birdcall Identification](https://www.kaggle.com/c/birdclef-2021)\n\n[1st place solution](https://www.kaggle.com/c/birdclef-2021/discussion/243927) given by @startjapan,\n - Nocall detector from Mel spectrograms\n - Multilabel classifier (backbone was resnest) for the Mel spectrogram\n - Feature engineering and it was analyzed by lightGBM\n\n[2nd place solution](https://www.kaggle.com/c/birdclef-2021/discussion/243463) given by @philippsinger,\n - Ensemble several CNNs, and using mixup\n - Mel spec transformation using [torchaudio](https://pytorch.org/audio/stable/index.html), CNN backbones from [Timm](https://github.com/rwightman/pytorch-image-models/)\n - Models: resnet34, tf_efficientnetv2_s_in21k, tf_efficientnetv2_m_in21k, eca_nfnet_l0\n\n[3rd place solution](https://www.kaggle.com/c/birdclef-2021/discussion/245708) given by @ludovick,\n - A backbone based on SED model (CNN)\n - Different architectures such as seresnet50, EfficientNetB2, EfficientNetB3, etc\n - Different CNNs has been trained such as SeResNet50, EfficientNetB2, EfficientNetB3, EfficientNetB4, EfficientNetB5, EfficientNetB6, EfficientNetB7\n\n[4th place solution](https://www.kaggle.com/c/birdclef-2021/discussion/243293) given by @tattaka,\n - Inference with global information for SED model\n - The logmelspectrograms were calculated using TorchAudio and normalized using the means and variances per image.\n - Create a pseudo label using the second method only for audio files without a secondary_label\n\n\n# [Freesound Audio Tagging 2019](https://www.kaggle.com/c/freesound-audio-tagging-2019/overview)\n![1](https://i.imgur.com/U6Z0oEx.png)\n[1st place solution](https://www.kaggle.com/c/freesound-audio-tagging-2019/discussion/95924) given by @romul0212,\n - CNN model with attention, skip connections, and auxiliary classifiers\n - SpecAugment, Mixup augmentations\n - Ensemble with MLP second-level model and geometric mean blending\n\n[2nd place solution](https://www.kaggle.com/c/freesound-audio-tagging-2019/discussion/97815) given by @action,\n - Feature engineering: log mel (441,64), global feature (128,12)\n - Audio clips are first trimmed of leading and trailing silence\n - Melspectrogram Layer, 9-layer CNN and highway + 1*1 conv\n\n[3rd place solution](https://www.kaggle.com/c/freesound-audio-tagging-2019/discussion/97926) given by @ddanevskyi,\n - 1d & 2d convolutions and 5-6 resnet blocks\n - Frequency encoding and Classification tasks\n - Data augmentation using MixUp and OR rule for mixing labels with 11 models ensemble\n\n[4th place solution](https://www.kaggle.com/c/freesound-audio-tagging-2019/discussion/96440) given by @osciiart,\n - Multitask learning and Semi-supervised learning (SSL) with noisy data\n - Models: ResNet34 with log-mel and EnvNet-v2 with waveform\n - Augmentations: Slicing, Mixup, Frequency masking, and Gain augmentation\n\n\n# [Bird audio detection](https://www.kaggle.com/c/bird-audio-detection)\n\n[1st place solution](https://www.kaggle.com/c/bird-audio-detection/discussion/202149) given by @umarzubair95,\n - Mel spectrograms from the training set (ffbird and wwbird datasets) and test set\n - CRNN approach, Covnet model, and Alexnet and LeNET models\n - Training with pseudo-label-approach\n\n\n# [Cornell Birdcall Identification](https://www.kaggle.com/c/birdsong-recognition)\n\n[1st place solution](https://www.kaggle.com/c/birdsong-recognition/discussion/183208) given by @taggatle,\n - 13 models with and without mixup\n - SpecAugmentation and AdamW with weight_decay 0.01\n - Cosine Annealing Scheduler with warmup\n\n[2nd place solution](https://www.kaggle.com/c/birdsong-recognition/discussion/183269) given by @vlomme\n - Pseudo-labels, predict classes on training files, and train using new labels, but the quality decreased slightly\n - Add a different sound without birds(rain, noise, conversations, etc.)\n - If there was a bird in the segment, I increased the probability of finding it in the entire file.\n\n[3rd place solution](https://www.kaggle.com/c/birdsong-recognition/discussion/183199) given by @theoviel,\n - Data Augmentations using Gaussian noise, Background noise, Modified Mixup\nand Improved cropping\n - Models: resnext50, resnext101 and resnest50\n - For the sites 1 and 2, the prediction of a given window is summed with those of the two neighbouring windows.", "votes": null }, { "id": "1561205", "postDate": "10/27/2021 12:19:36", "content": "

Hey All,

\n

Thank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey https://forms.gle/QP9L16niPexozyhu5.

\n

Thank you all,

\n

Regards,
\nChris

", "rawMarkdown": "Hey All,\n\nThank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey https://forms.gle/QP9L16niPexozyhu5.\n\nThank you all,\n\nRegards,\nChris", "votes": null } ], "comments": [ { "id": 1561205, "author_name": "zerafachris", "author_url": "", "post_date": "10/27/2021 12:19:36", "content": "

Hey All,

\n

Thank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey https://forms.gle/QP9L16niPexozyhu5.

\n

Thank you all,

\n

Regards,
\nChris

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1466240": "`Previous top solutions of the similar Competitions`\n#[BirdCLEF 2021 - Birdcall Identification](https://www.kaggle.com/c/birdclef-2021)\n\n[1st place solution](https://www.kaggle.com/c/birdclef-2021/discussion/243927) given by @startjapan,\n - Nocall detector from Mel spectrograms\n - Multilabel classifier (backbone was resnest) for the Mel spectrogram\n - Feature engineering and it was analyzed by lightGBM\n\n[2nd place solution](https://www.kaggle.com/c/birdclef-2021/discussion/243463) given by @philippsinger,\n - Ensemble several CNNs, and using mixup\n - Mel spec transformation using [torchaudio](https://pytorch.org/audio/stable/index.html), CNN backbones from [Timm](https://github.com/rwightman/pytorch-image-models/)\n - Models: resnet34, tf_efficientnetv2_s_in21k, tf_efficientnetv2_m_in21k, eca_nfnet_l0\n\n[3rd place solution](https://www.kaggle.com/c/birdclef-2021/discussion/245708) given by @ludovick,\n - A backbone based on SED model (CNN)\n - Different architectures such as seresnet50, EfficientNetB2, EfficientNetB3, etc\n - Different CNNs has been trained such as SeResNet50, EfficientNetB2, EfficientNetB3, EfficientNetB4, EfficientNetB5, EfficientNetB6, EfficientNetB7\n\n[4th place solution](https://www.kaggle.com/c/birdclef-2021/discussion/243293) given by @tattaka,\n - Inference with global information for SED model\n - The logmelspectrograms were calculated using TorchAudio and normalized using the means and variances per image.\n - Create a pseudo label using the second method only for audio files without a secondary_label\n\n\n# [Freesound Audio Tagging 2019](https://www.kaggle.com/c/freesound-audio-tagging-2019/overview)\n![1](https://i.imgur.com/U6Z0oEx.png)\n[1st place solution](https://www.kaggle.com/c/freesound-audio-tagging-2019/discussion/95924) given by @romul0212,\n - CNN model with attention, skip connections, and auxiliary classifiers\n - SpecAugment, Mixup augmentations\n - Ensemble with MLP second-level model and geometric mean blending\n\n[2nd place solution](https://www.kaggle.com/c/freesound-audio-tagging-2019/discussion/97815) given by @action,\n - Feature engineering: log mel (441,64), global feature (128,12)\n - Audio clips are first trimmed of leading and trailing silence\n - Melspectrogram Layer, 9-layer CNN and highway + 1*1 conv\n\n[3rd place solution](https://www.kaggle.com/c/freesound-audio-tagging-2019/discussion/97926) given by @ddanevskyi,\n - 1d & 2d convolutions and 5-6 resnet blocks\n - Frequency encoding and Classification tasks\n - Data augmentation using MixUp and OR rule for mixing labels with 11 models ensemble\n\n[4th place solution](https://www.kaggle.com/c/freesound-audio-tagging-2019/discussion/96440) given by @osciiart,\n - Multitask learning and Semi-supervised learning (SSL) with noisy data\n - Models: ResNet34 with log-mel and EnvNet-v2 with waveform\n - Augmentations: Slicing, Mixup, Frequency masking, and Gain augmentation\n\n\n# [Bird audio detection](https://www.kaggle.com/c/bird-audio-detection)\n\n[1st place solution](https://www.kaggle.com/c/bird-audio-detection/discussion/202149) given by @umarzubair95,\n - Mel spectrograms from the training set (ffbird and wwbird datasets) and test set\n - CRNN approach, Covnet model, and Alexnet and LeNET models\n - Training with pseudo-label-approach\n\n\n# [Cornell Birdcall Identification](https://www.kaggle.com/c/birdsong-recognition)\n\n[1st place solution](https://www.kaggle.com/c/birdsong-recognition/discussion/183208) given by @taggatle,\n - 13 models with and without mixup\n - SpecAugmentation and AdamW with weight_decay 0.01\n - Cosine Annealing Scheduler with warmup\n\n[2nd place solution](https://www.kaggle.com/c/birdsong-recognition/discussion/183269) given by @vlomme\n - Pseudo-labels, predict classes on training files, and train using new labels, but the quality decreased slightly\n - Add a different sound without birds(rain, noise, conversations, etc.)\n - If there was a bird in the segment, I increased the probability of finding it in the entire file.\n\n[3rd place solution](https://www.kaggle.com/c/birdsong-recognition/discussion/183199) given by @theoviel,\n - Data Augmentations using Gaussian noise, Background noise, Modified Mixup\nand Improved cropping\n - Models: resnext50, resnext101 and resnest50\n - For the sites 1 and 2, the prediction of a given window is summed with those of the two neighbouring windows.", "1561205": "Hey All,\n\nThank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey https://forms.gle/QP9L16niPexozyhu5.\n\nThank you all,\n\nRegards,\nChris" }, "source": "meta" }