{
  "id": 251549,
  "title": "[CV vs LB]",
  "url": "/competitions/g2net-gravitational-wave-detection/discussion/251549",
  "author_name": "Tahsin Mostafiz",
  "post_date": "2021-07-07T19:00:58.365000",
  "votes": 26,
  "comment_count": 15,
  "views": 0,
  "content": "<p>My Baseline:<br>\nStratified 5 fold data<br>\nApplied Q-Transform based on <a href=\"https://www.kaggle.com/yasufuminakama/g2net-efficientnet-b7-baseline-training?scriptVersionId=67408619\" target=\"_blank\">this</a> notebook<br>\nModel: EffB0<br>\nEpoch: 5<br>\nAug: None<br>\nLoss: BCEWithLogitsLoss<br>\nSingle Fold<br>\nCV: 0.8568<br>\nLB: 0.862</p>",
  "messages": [
    {
      "id": 1380061,
      "postDate": "2021-07-07T19:00:58.367Z",
      "content": "<p>My Baseline:<br>\nStratified 5 fold data<br>\nApplied Q-Transform based on <a href=\"https://www.kaggle.com/yasufuminakama/g2net-efficientnet-b7-baseline-training?scriptVersionId=67408619\" target=\"_blank\">this</a> notebook<br>\nModel: EffB0<br>\nEpoch: 5<br>\nAug: None<br>\nLoss: BCEWithLogitsLoss<br>\nSingle Fold<br>\nCV: 0.8568<br>\nLB: 0.862</p>",
      "rawMarkdown": "My Baseline:\nStratified 5 fold data\nApplied Q-Transform based on [this](https://www.kaggle.com/yasufuminakama/g2net-efficientnet-b7-baseline-training?scriptVersionId=67408619) notebook\nModel: EffB0\nEpoch: 5\nAug: None\nLoss: BCEWithLogitsLoss\nSingle Fold\nCV: 0.8568\nLB: 0.862",
      "votes": 26
    },
    {
      "id": 1380911,
      "postDate": "2021-07-08T12:35:45.973Z",
      "content": "<p>My Baseline:<br>\nStratified 5 fold data<br>\nApplied Q-Transform <br>\nModel: EffB0<br>\nEpoch: 3<br>\nAug: None<br>\nLoss: BCEWithLogitsLoss<br>\n5 folds<br>\nCV: 0.8634<br>\nLB: 0.866</p>",
      "rawMarkdown": "My Baseline:\nStratified 5 fold data\nApplied Q-Transform \nModel: EffB0\nEpoch: 3\nAug: None\nLoss: BCEWithLogitsLoss\n5 folds\nCV: 0.8634\nLB: 0.866",
      "votes": 6
    },
    {
      "id": 1461696,
      "postDate": "2021-08-09T14:21:11.840Z",
      "content": "<p>My Baseline was like that </p>\n<p>Stratified 5<br>\nApplied Q-Transform<br>\nModel: EffB0<br>\nEpoch: 3<br>\nAug: None<br>\nLoss: BCEWithLogitsLoss<br>\n5 folds<br>\nCV: 0.873<br>\nLB: 0.874</p>",
      "rawMarkdown": "My Baseline was like that \n\nStratified 5\nApplied Q-Transform\nModel: EffB0\nEpoch: 3\nAug: None\nLoss: BCEWithLogitsLoss\n5 folds\nCV: 0.873\nLB: 0.874",
      "votes": 4
    },
    {
      "id": 1380286,
      "postDate": "2021-07-08T01:35:53.703Z",
      "content": "<p>My Baseline:<br>\nStratified 5 fold data<br>\nApplied Q-Transform based on this notebook(<strong>3channel</strong>)<br>\nModel: EffB0<br>\nEpoch: 5<br>\nAug: None<br>\nLoss: BCEWithLogitsLoss<br>\nSingle Fold<br>\nCV: 0.8561<br>\nLB: 0.862</p>",
      "rawMarkdown": "My Baseline:\nStratified 5 fold data\nApplied Q-Transform based on this notebook(**3channel**)\nModel: EffB0\nEpoch: 5\nAug: None\nLoss: BCEWithLogitsLoss\nSingle Fold\nCV: 0.8561\nLB: 0.862",
      "votes": 4,
      "replies": [
        {
          "id": 1381446,
          "postDate": "2021-07-09T02:44:32.323Z",
          "content": "<p>update:<br>\nMy Baseline:<br>\nStratified 5 fold data<br>\nApplied Q-Transform based on this notebook(3channel)<br>\nModel: EffV2_B1<br>\nEpoch: 5<br>\nAug: cutmix<br>\nLoss: BCEWithLogitsLoss<br>\nSingle Fold<br>\nCV: 0.8609<br>\nLB: 0.865</p>\n<p>update:<br>\nMy Baseline:<br>\nStratified 5 fold data<br>\nApplied Q-Transform based on this notebook(3channel)<br>\nModel: EffV2_B1<br>\nEpoch: 3<br>\nAug: cutmix<br>\nLoss: BCEWithLogitsLoss<br>\nimage_size:Larger image<br>\nSingle Fold<br>\nCV: 0.8639<br>\nLB: 0.867</p>\n<p>update:<br>\nMy Baseline:<br>\nStratified 5 fold data<br>\nApplied Q-Transform based on this notebook(3channel)<br>\nModel: EffV2_B1<br>\nEpoch: 3<br>\nAug: cutmix<br>\nLoss: BCEWithLogitsLoss<br>\nimage_size:Larger Larger  image<br>\nSingle Fold<br>\nCV: 0.86566<br>\nLB: 0.870</p>",
          "rawMarkdown": "update:\nMy Baseline:\nStratified 5 fold data\nApplied Q-Transform based on this notebook(3channel)\nModel: EffV2_B1\nEpoch: 5\nAug: cutmix\nLoss: BCEWithLogitsLoss\nSingle Fold\nCV: 0.8609\nLB: 0.865\n\nupdate:\nMy Baseline:\nStratified 5 fold data\nApplied Q-Transform based on this notebook(3channel)\nModel: EffV2_B1\nEpoch: 3\nAug: cutmix\nLoss: BCEWithLogitsLoss\nimage_size:Larger image\nSingle Fold\nCV: 0.8639\nLB: 0.867\n\nupdate:\nMy Baseline:\nStratified 5 fold data\nApplied Q-Transform based on this notebook(3channel)\nModel: EffV2_B1\nEpoch: 3\nAug: cutmix\nLoss: BCEWithLogitsLoss\nimage_size:Larger Larger  image\nSingle Fold\nCV: 0.86566\nLB: 0.870",
          "votes": 4
        },
        {
          "id": 1404548,
          "postDate": "2021-07-30T03:02:29.560Z",
          "content": "<p>Can you please the notebook from which you used q-transforms ? </p>",
          "rawMarkdown": "Can you please the notebook from which you used q-transforms ? "
        }
      ]
    },
    {
      "id": 1484773,
      "postDate": "2021-08-21T15:06:16.510Z",
      "content": "<p>Single Model 5 fold score…<br>\nCV-87.1<br>\nLB-87.4</p>",
      "rawMarkdown": "Single Model 5 fold score...\nCV-87.1\nLB-87.4",
      "votes": 1
    },
    {
      "id": 1381745,
      "postDate": "2021-07-09T07:43:38.627Z",
      "content": "<p>My Baseline:<br>\nStratified 5 fold data<br>\nUsed Q Transform<br>\nEpoch: 5<br>\nAug: None<br>\nLoss: BCEWithLogitsLoss<br>\nSingle Fold<br>\nCV: 0.8605<br>\nLB: 0.863</p>",
      "rawMarkdown": "My Baseline:\nStratified 5 fold data\nUsed Q Transform\nEpoch: 5\nAug: None\nLoss: BCEWithLogitsLoss\nSingle Fold\nCV: 0.8605\nLB: 0.863",
      "votes": 1
    },
    {
      "id": 1461127,
      "postDate": "2021-08-09T08:21:06.137Z",
      "content": "<p>My Baseline:<br>\nStratified 5 fold data<br>\nQ-Transform &amp; bandpass filter<br>\nModel: ResNet50d<br>\nAug: None<br>\nLoss: BCEWithLogitsLoss<br>\n5 folds<br>\nCV: 0.8659<br>\nLB: 0.871</p>",
      "rawMarkdown": "My Baseline:\nStratified 5 fold data\nQ-Transform & bandpass filter\nModel: ResNet50d\nAug: None\nLoss: BCEWithLogitsLoss\n5 folds\nCV: 0.8659\nLB: 0.871",
      "votes": 2
    },
    {
      "id": 1561182,
      "postDate": "2021-10-27T12:15:15.030Z",
      "content": "<p>Hey All,</p>\n<p>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 <a href=\"https://forms.gle/QP9L16niPexozyhu5\" target=\"_blank\">https://forms.gle/QP9L16niPexozyhu5</a>.</p>\n<p>Thank you all,</p>\n<p>Regards,<br>\nChris</p>",
      "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"
    },
    {
      "id": 1486457,
      "postDate": "2021-08-23T01:54:38.940Z",
      "content": "<p>single Model CWT 5 fold score:-<br>\nCV- 0.872<br>\nLB-0.874</p>",
      "rawMarkdown": "single Model CWT 5 fold score:-\nCV- 0.872\nLB-0.874"
    },
    {
      "id": 1486147,
      "postDate": "2021-08-22T17:36:03.327Z",
      "content": "<p>Another thread with same theme: <a href=\"https://www.kaggle.com/c/g2net-gravitational-wave-detection/discussion/264380\" target=\"_blank\">https://www.kaggle.com/c/g2net-gravitational-wave-detection/discussion/264380</a></p>",
      "rawMarkdown": "Another thread with same theme: https://www.kaggle.com/c/g2net-gravitational-wave-detection/discussion/264380"
    },
    {
      "id": 1470455,
      "postDate": "2021-08-13T14:21:46.347Z",
      "content": "<p><a href=\"https://www.kaggle.com/tahsin\" target=\"_blank\">@tahsin</a>  could u elaborate when you said applied CQT on this notebook.<br>\nThe notebook you shown is already doing same </p>",
      "rawMarkdown": "@tahsin  could u elaborate when you said applied CQT on this notebook.\nThe notebook you shown is already doing same \n"
    },
    {
      "id": 1463174,
      "postDate": "2021-08-10T05:56:03.150Z",
      "content": "<h4>My Baseline</h4>\n<table>\n<thead>\n<tr>\n<th><strong>Type</strong></th>\n<th><strong>Method Used</strong></th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>Strategy</td>\n<td>4 Fold CV</td>\n</tr>\n<tr>\n<td>Model</td>\n<td>EfficientNet B7</td>\n</tr>\n<tr>\n<td>Dataset</td>\n<td>CQT (256x256)</td>\n</tr>\n<tr>\n<td>Augmentation</td>\n<td>Mixup</td>\n</tr>\n<tr>\n<td>Epochs</td>\n<td>20</td>\n</tr>\n<tr>\n<td>LB</td>\n<td>0.862</td>\n</tr>\n<tr>\n<td>CV</td>\n<td>0.868</td>\n</tr>\n</tbody>\n</table>",
      "rawMarkdown": "#### My Baseline\n\n| **Type** | **Method Used** |\n| :---: | :---: |\n| Strategy | 4 Fold CV |\n| Model | EfficientNet B7 |\n| Dataset | CQT (256x256) |\n| Augmentation | Mixup |\n| Epochs | 20 |\n| LB | 0.862 |\n| CV | 0.868 |"
    },
    {
      "id": 1381593,
      "postDate": "2021-07-09T06:25:40.213Z",
      "content": "<p>Thanks for sharing <a href=\"https://www.kaggle.com/tahsin\" target=\"_blank\">@tahsin</a> , may I ask why all of you trained with so less epochs(3 ~ 5), is it because the train samples are too much(560k)? And how long will your one epoch spend, with which GPU? Thanks.</p>",
      "rawMarkdown": "Thanks for sharing @tahsin , may I ask why all of you trained with so less epochs(3 ~ 5), is it because the train samples are too much(560k)? And how long will your one epoch spend, with which GPU? Thanks."
    },
    {
      "id": 1461816,
      "postDate": "2021-08-09T15:17:43.293Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 1380911,
      "author_name": "CPMP",
      "author_url": "",
      "post_date": "2021-07-08T12:35:45.973000",
      "content": "<p>My Baseline:<br>\nStratified 5 fold data<br>\nApplied Q-Transform <br>\nModel: EffB0<br>\nEpoch: 3<br>\nAug: None<br>\nLoss: BCEWithLogitsLoss<br>\n5 folds<br>\nCV: 0.8634<br>\nLB: 0.866</p>",
      "votes": 6,
      "replies": []
    },
    {
      "id": 1461696,
      "author_name": "stallone",
      "author_url": "",
      "post_date": "2021-08-09T14:21:11.840000",
      "content": "<p>My Baseline was like that </p>\n<p>Stratified 5<br>\nApplied Q-Transform<br>\nModel: EffB0<br>\nEpoch: 3<br>\nAug: None<br>\nLoss: BCEWithLogitsLoss<br>\n5 folds<br>\nCV: 0.873<br>\nLB: 0.874</p>",
      "votes": 4,
      "replies": []
    },
    {
      "id": 1380286,
      "author_name": "zhangeng",
      "author_url": "",
      "post_date": "2021-07-08T01:35:53.703000",
      "content": "<p>My Baseline:<br>\nStratified 5 fold data<br>\nApplied Q-Transform based on this notebook(<strong>3channel</strong>)<br>\nModel: EffB0<br>\nEpoch: 5<br>\nAug: None<br>\nLoss: BCEWithLogitsLoss<br>\nSingle Fold<br>\nCV: 0.8561<br>\nLB: 0.862</p>",
      "votes": 4,
      "replies": [
        {
          "id": 1381446,
          "author_name": "zhangeng",
          "author_url": "",
          "post_date": "2021-07-09T02:44:32.323000",
          "content": "<p>update:<br>\nMy Baseline:<br>\nStratified 5 fold data<br>\nApplied Q-Transform based on this notebook(3channel)<br>\nModel: EffV2_B1<br>\nEpoch: 5<br>\nAug: cutmix<br>\nLoss: BCEWithLogitsLoss<br>\nSingle Fold<br>\nCV: 0.8609<br>\nLB: 0.865</p>\n<p>update:<br>\nMy Baseline:<br>\nStratified 5 fold data<br>\nApplied Q-Transform based on this notebook(3channel)<br>\nModel: EffV2_B1<br>\nEpoch: 3<br>\nAug: cutmix<br>\nLoss: BCEWithLogitsLoss<br>\nimage_size:Larger image<br>\nSingle Fold<br>\nCV: 0.8639<br>\nLB: 0.867</p>\n<p>update:<br>\nMy Baseline:<br>\nStratified 5 fold data<br>\nApplied Q-Transform based on this notebook(3channel)<br>\nModel: EffV2_B1<br>\nEpoch: 3<br>\nAug: cutmix<br>\nLoss: BCEWithLogitsLoss<br>\nimage_size:Larger Larger  image<br>\nSingle Fold<br>\nCV: 0.86566<br>\nLB: 0.870</p>",
          "votes": 4,
          "replies": []
        },
        {
          "id": 1404548,
          "author_name": "Mithil Salunkhe",
          "author_url": "",
          "post_date": "2021-07-30T03:02:29.560000",
          "content": "<p>Can you please the notebook from which you used q-transforms ? </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1484773,
      "author_name": "Jaideep",
      "author_url": "",
      "post_date": "2021-08-21T15:06:16.510000",
      "content": "<p>Single Model 5 fold score…<br>\nCV-87.1<br>\nLB-87.4</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 1381745,
      "author_name": "Athar Sayed",
      "author_url": "",
      "post_date": "2021-07-09T07:43:38.627000",
      "content": "<p>My Baseline:<br>\nStratified 5 fold data<br>\nUsed Q Transform<br>\nEpoch: 5<br>\nAug: None<br>\nLoss: BCEWithLogitsLoss<br>\nSingle Fold<br>\nCV: 0.8605<br>\nLB: 0.863</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 1461127,
      "author_name": "datasaurus",
      "author_url": "",
      "post_date": "2021-08-09T08:21:06.137000",
      "content": "<p>My Baseline:<br>\nStratified 5 fold data<br>\nQ-Transform &amp; bandpass filter<br>\nModel: ResNet50d<br>\nAug: None<br>\nLoss: BCEWithLogitsLoss<br>\n5 folds<br>\nCV: 0.8659<br>\nLB: 0.871</p>",
      "votes": 2,
      "replies": []
    },
    {
      "id": 1561182,
      "author_name": "ChristopherZerafa",
      "author_url": "",
      "post_date": "2021-10-27T12:15:15.030000",
      "content": "<p>Hey All,</p>\n<p>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 <a href=\"https://forms.gle/QP9L16niPexozyhu5\" target=\"_blank\">https://forms.gle/QP9L16niPexozyhu5</a>.</p>\n<p>Thank you all,</p>\n<p>Regards,<br>\nChris</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1486457,
      "author_name": "shiroe",
      "author_url": "",
      "post_date": "2021-08-23T01:54:38.940000",
      "content": "<p>single Model CWT 5 fold score:-<br>\nCV- 0.872<br>\nLB-0.874</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1486147,
      "author_name": "Araik Tamazian",
      "author_url": "",
      "post_date": "2021-08-22T17:36:03.327000",
      "content": "<p>Another thread with same theme: <a href=\"https://www.kaggle.com/c/g2net-gravitational-wave-detection/discussion/264380\" target=\"_blank\">https://www.kaggle.com/c/g2net-gravitational-wave-detection/discussion/264380</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1470455,
      "author_name": "Jaideep",
      "author_url": "",
      "post_date": "2021-08-13T14:21:46.347000",
      "content": "<p><a href=\"https://www.kaggle.com/tahsin\" target=\"_blank\">@tahsin</a>  could u elaborate when you said applied CQT on this notebook.<br>\nThe notebook you shown is already doing same </p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1463174,
      "author_name": "Saurav Maheshkar ☕️",
      "author_url": "",
      "post_date": "2021-08-10T05:56:03.150000",
      "content": "<h4>My Baseline</h4>\n<table>\n<thead>\n<tr>\n<th><strong>Type</strong></th>\n<th><strong>Method Used</strong></th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>Strategy</td>\n<td>4 Fold CV</td>\n</tr>\n<tr>\n<td>Model</td>\n<td>EfficientNet B7</td>\n</tr>\n<tr>\n<td>Dataset</td>\n<td>CQT (256x256)</td>\n</tr>\n<tr>\n<td>Augmentation</td>\n<td>Mixup</td>\n</tr>\n<tr>\n<td>Epochs</td>\n<td>20</td>\n</tr>\n<tr>\n<td>LB</td>\n<td>0.862</td>\n</tr>\n<tr>\n<td>CV</td>\n<td>0.868</td>\n</tr>\n</tbody>\n</table>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1381593,
      "author_name": "Hao",
      "author_url": "",
      "post_date": "2021-07-09T06:25:40.213000",
      "content": "<p>Thanks for sharing <a href=\"https://www.kaggle.com/tahsin\" target=\"_blank\">@tahsin</a> , may I ask why all of you trained with so less epochs(3 ~ 5), is it because the train samples are too much(560k)? And how long will your one epoch spend, with which GPU? Thanks.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1461816,
      "author_name": "",
      "author_url": "",
      "post_date": "2021-08-09T15:17:43.293000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1380061": "My Baseline:\nStratified 5 fold data\nApplied Q-Transform based on [this](https://www.kaggle.com/yasufuminakama/g2net-efficientnet-b7-baseline-training?scriptVersionId=67408619) notebook\nModel: EffB0\nEpoch: 5\nAug: None\nLoss: BCEWithLogitsLoss\nSingle Fold\nCV: 0.8568\nLB: 0.862",
    "1380911": "My Baseline:\nStratified 5 fold data\nApplied Q-Transform \nModel: EffB0\nEpoch: 3\nAug: None\nLoss: BCEWithLogitsLoss\n5 folds\nCV: 0.8634\nLB: 0.866",
    "1461696": "My Baseline was like that \n\nStratified 5\nApplied Q-Transform\nModel: EffB0\nEpoch: 3\nAug: None\nLoss: BCEWithLogitsLoss\n5 folds\nCV: 0.873\nLB: 0.874",
    "1380286": "My Baseline:\nStratified 5 fold data\nApplied Q-Transform based on this notebook(**3channel**)\nModel: EffB0\nEpoch: 5\nAug: None\nLoss: BCEWithLogitsLoss\nSingle Fold\nCV: 0.8561\nLB: 0.862",
    "1484773": "Single Model 5 fold score...\nCV-87.1\nLB-87.4",
    "1381745": "My Baseline:\nStratified 5 fold data\nUsed Q Transform\nEpoch: 5\nAug: None\nLoss: BCEWithLogitsLoss\nSingle Fold\nCV: 0.8605\nLB: 0.863",
    "1461127": "My Baseline:\nStratified 5 fold data\nQ-Transform & bandpass filter\nModel: ResNet50d\nAug: None\nLoss: BCEWithLogitsLoss\n5 folds\nCV: 0.8659\nLB: 0.871",
    "1561182": "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",
    "1486457": "single Model CWT 5 fold score:-\nCV- 0.872\nLB-0.874",
    "1486147": "Another thread with same theme: https://www.kaggle.com/c/g2net-gravitational-wave-detection/discussion/264380",
    "1470455": "@tahsin  could u elaborate when you said applied CQT on this notebook.\nThe notebook you shown is already doing same \n",
    "1463174": "#### My Baseline\n\n| **Type** | **Method Used** |\n| :---: | :---: |\n| Strategy | 4 Fold CV |\n| Model | EfficientNet B7 |\n| Dataset | CQT (256x256) |\n| Augmentation | Mixup |\n| Epochs | 20 |\n| LB | 0.862 |\n| CV | 0.868 |",
    "1381593": "Thanks for sharing @tahsin , may I ask why all of you trained with so less epochs(3 ~ 5), is it because the train samples are too much(560k)? And how long will your one epoch spend, with which GPU? Thanks.",
    "1461816": ""
  }
}