{
  "id": 238037,
  "title": " best single model 94.8 , Congrats to survivors , for rest  learning ",
  "url": "/competitions/hubmap-kidney-segmentation/discussion/238037",
  "author_name": "Jaideep",
  "post_date": "2021-05-11T03:09:04.163000",
  "votes": 6,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I heartily congratulate those who managed to stay afloat in medal zone inspite of heavy lb storm .<br>\nWe selected one submission scoring 93 on public and 94 on private and other  scoring 94 on public and 93 6 on private  . Model used in highest public score model was unet regnetx064 .regenetx064 was best encoder model for us .</p>\n<p>But our second submission couldn't save us much and we were  one among those severely affected thanks to not able to select best single best model  submission that scored 94.8   ( deeplabv3plus backbone regnetx 064 no hard core pl training just included hl  d4 by zhao )  on private and 93.5 on public .</p>\n<p><a href=\"https://www.kaggle.com/jaideepvalani/deepflash-updated-raster?scriptVersionId=62377281\" target=\"_blank\">https://www.kaggle.com/jaideepvalani/deepflash-updated-raster?scriptVersionId=62377281</a><br>\n94.8<br>\n<strong>Update:</strong><br>\nEnsemble<br>\n<a href=\"https://www.kaggle.com/jaideepvalani/deepflash-updated-raster?scriptVersionId=62573188\" target=\"_blank\">https://www.kaggle.com/jaideepvalani/deepflash-updated-raster?scriptVersionId=62573188</a> -94.9<br>\nMost important was selection in this competition I think.But overall  objective of this competition  that is to detect FTu with such high accuracy was pretty much met by many kagglers so they should feel happy with that .and taker better selection next time .<br>\nKeep trying see u in other competion. </p>\n<p>Regards <br>\nJaideep</p>",
  "messages": [
    {
      "id": 1301335,
      "postDate": "2021-05-11T03:09:04.163Z",
      "content": "<p>I heartily congratulate those who managed to stay afloat in medal zone inspite of heavy lb storm .<br>\nWe selected one submission scoring 93 on public and 94 on private and other  scoring 94 on public and 93 6 on private  . Model used in highest public score model was unet regnetx064 .regenetx064 was best encoder model for us .</p>\n<p>But our second submission couldn't save us much and we were  one among those severely affected thanks to not able to select best single best model  submission that scored 94.8   ( deeplabv3plus backbone regnetx 064 no hard core pl training just included hl  d4 by zhao )  on private and 93.5 on public .</p>\n<p><a href=\"https://www.kaggle.com/jaideepvalani/deepflash-updated-raster?scriptVersionId=62377281\" target=\"_blank\">https://www.kaggle.com/jaideepvalani/deepflash-updated-raster?scriptVersionId=62377281</a><br>\n94.8<br>\n<strong>Update:</strong><br>\nEnsemble<br>\n<a href=\"https://www.kaggle.com/jaideepvalani/deepflash-updated-raster?scriptVersionId=62573188\" target=\"_blank\">https://www.kaggle.com/jaideepvalani/deepflash-updated-raster?scriptVersionId=62573188</a> -94.9<br>\nMost important was selection in this competition I think.But overall  objective of this competition  that is to detect FTu with such high accuracy was pretty much met by many kagglers so they should feel happy with that .and taker better selection next time .<br>\nKeep trying see u in other competion. </p>\n<p>Regards <br>\nJaideep</p>",
      "rawMarkdown": "I heartily congratulate those who managed to stay afloat in medal zone inspite of heavy lb storm .\nWe selected one submission scoring 93 on public and 94 on private and other  scoring 94 on public and 93 6 on private  . Model used in highest public score model was unet regnetx064 .regenetx064 was best encoder model for us .\n\nBut our second submission couldn't save us much and we were  one among those severely affected thanks to not able to select best single best model  submission that scored 94.8   ( deeplabv3plus backbone regnetx 064 no hard core pl training just included hl  d4 by zhao )  on private and 93.5 on public .\n\nhttps://www.kaggle.com/jaideepvalani/deepflash-updated-raster?scriptVersionId=62377281\n94.8\n**Update:**\nEnsemble\nhttps://www.kaggle.com/jaideepvalani/deepflash-updated-raster?scriptVersionId=62573188 -94.9\nMost important was selection in this competition I think.But overall  objective of this competition  that is to detect FTu with such high accuracy was pretty much met by many kagglers so they should feel happy with that .and taker better selection next time .\nKeep trying see u in other competion. \n\n\nRegards \nJaideep\n",
      "votes": 6
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1301335": "I heartily congratulate those who managed to stay afloat in medal zone inspite of heavy lb storm .\nWe selected one submission scoring 93 on public and 94 on private and other  scoring 94 on public and 93 6 on private  . Model used in highest public score model was unet regnetx064 .regenetx064 was best encoder model for us .\n\nBut our second submission couldn't save us much and we were  one among those severely affected thanks to not able to select best single best model  submission that scored 94.8   ( deeplabv3plus backbone regnetx 064 no hard core pl training just included hl  d4 by zhao )  on private and 93.5 on public .\n\nhttps://www.kaggle.com/jaideepvalani/deepflash-updated-raster?scriptVersionId=62377281\n94.8\n**Update:**\nEnsemble\nhttps://www.kaggle.com/jaideepvalani/deepflash-updated-raster?scriptVersionId=62573188 -94.9\nMost important was selection in this competition I think.But overall  objective of this competition  that is to detect FTu with such high accuracy was pretty much met by many kagglers so they should feel happy with that .and taker better selection next time .\nKeep trying see u in other competion. \n\n\nRegards \nJaideep\n"
  }
}