{
  "id": 198244,
  "title": "Ideas for Further Improving Public Kernels",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/198244",
  "author_name": "",
  "post_date": "2020-11-20T11:14:51.075435600Z",
  "votes": 51,
  "comment_count": 8,
  "views": 0,
  "content": "<ol>\n<li>Try with different pretrained efficientnet/se-resnext models</li>\n<li>Try with different image sizes</li>\n<li>Try with different augmentations</li>\n<li>Finetune lr scheduling </li>\n<li>Try a different optimizer: Adabelief (?)</li>\n<li>Train 5-fold models and save the oof predictions for point 7.</li>\n<li>Blending\\Stacking</li>\n<li>Cutmix with soft labels</li>\n</ol>\n<p>Anyone has more ideas to share :) ?</p>",
  "messages": [
    {
      "id": "1084768",
      "postDate": "11/20/2020 11:14:51",
      "content": "<ol>\n<li>Try with different pretrained efficientnet/se-resnext models</li>\n<li>Try with different image sizes</li>\n<li>Try with different augmentations</li>\n<li>Finetune lr scheduling </li>\n<li>Try a different optimizer: Adabelief (?)</li>\n<li>Train 5-fold models and save the oof predictions for point 7.</li>\n<li>Blending\\Stacking</li>\n<li>Cutmix with soft labels</li>\n</ol>\n<p>Anyone has more ideas to share :) ?</p>",
      "rawMarkdown": "1.  Try with different pretrained efficientnet/se-resnext models\n2. Try with different image sizes\n3. Try with different augmentations\n4. Finetune lr scheduling \n5. Try a different optimizer: Adabelief (?)\n6. Train 5-fold models and save the oof predictions for point 7.\n7. Blending\\Stacking\n8. Cutmix with soft labels\n\nAnyone has more ideas to share :) ?",
      "votes": null
    },
    {
      "id": "1087062",
      "postDate": "11/22/2020 10:01:15",
      "content": "<p>Nice tips. What about using an external data set! Is there any resource? </p>",
      "rawMarkdown": "Nice tips. What about using an external data set! Is there any resource?",
      "votes": null
    },
    {
      "id": "1087100",
      "postDate": "11/22/2020 10:51:28",
      "content": "<p>Good suggestion!</p>",
      "rawMarkdown": "Good suggestion!",
      "votes": null
    },
    {
      "id": "1088942",
      "postDate": "11/24/2020 04:38:53",
      "content": "<p>Some of the papers suggested on other posts suggested svm instead of softmax in the final layer </p>",
      "rawMarkdown": "Some of the papers suggested on other posts suggested svm instead of softmax in the final layer",
      "votes": null
    },
    {
      "id": "1089042",
      "postDate": "11/24/2020 06:49:45",
      "content": "<p>Can you share the link of these papers? Thanks</p>",
      "rawMarkdown": "Can you share the link of these papers? Thanks",
      "votes": null
    },
    {
      "id": "1089055",
      "postDate": "11/24/2020 07:04:57",
      "content": "<p><a href=\"https://www.kaggle.com/c/cassava-leaf-disease-classification/discussion/198146\" target=\"_blank\">https://www.kaggle.com/c/cassava-leaf-disease-classification/discussion/198146</a></p>\n<p>iCassava 2019 Fine-Grained Visual Categorization Challenge<br>\nUsing Transfer Learning for Image-Based Cassava Disease Detection</p>",
      "rawMarkdown": "https://www.kaggle.com/c/cassava-leaf-disease-classification/discussion/198146\n\niCassava 2019 Fine-Grained Visual Categorization Challenge\nUsing Transfer Learning for Image-Based Cassava Disease Detection",
      "votes": null
    },
    {
      "id": "1090336",
      "postDate": "11/25/2020 09:08:41",
      "content": "<p>anyone applied label smoothing??</p>",
      "rawMarkdown": "anyone applied label smoothing??",
      "votes": null
    },
    {
      "id": "1125417",
      "postDate": "12/24/2020 17:41:42",
      "content": "<p><a href=\"https://www.kaggle.com/c/cassava-leaf-disease-classification/discussion/205491\" target=\"_blank\">7 things that did not worked so far</a><br>\nhere the author says, external dataset didn't help him that much.</p>",
      "rawMarkdown": "[7 things that did not worked so far](https://www.kaggle.com/c/cassava-leaf-disease-classification/discussion/205491)\nhere the author says, external dataset didn't help him that much.",
      "votes": null
    },
    {
      "id": "1144187",
      "postDate": "01/08/2021 09:48:31",
      "content": "<p><a href=\"https://www.kaggle.com/khyeh0719\" target=\"_blank\">@khyeh0719</a> I know what 5 folds and oof means but what does this means save oof predictions for point 7 can you please explain .</p>",
      "rawMarkdown": "khyeh0719 I know what 5 folds and oof means but what does this means save oof predictions for point 7 can you please explain .",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1087062,
      "author_name": "ipythonx",
      "author_url": "",
      "post_date": "11/22/2020 10:01:15",
      "content": "<p>Nice tips. What about using an external data set! Is there any resource? </p>",
      "votes": null,
      "replies": [
        {
          "id": 1087100,
          "author_name": "khyeh0719",
          "author_url": "",
          "post_date": "11/22/2020 10:51:28",
          "content": "<p>Good suggestion!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1125417,
          "author_name": "jabertuhin",
          "author_url": "",
          "post_date": "12/24/2020 17:41:42",
          "content": "<p><a href=\"https://www.kaggle.com/c/cassava-leaf-disease-classification/discussion/205491\" target=\"_blank\">7 things that did not worked so far</a><br>\nhere the author says, external dataset didn't help him that much.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1088942,
      "author_name": "venkat555",
      "author_url": "",
      "post_date": "11/24/2020 04:38:53",
      "content": "<p>Some of the papers suggested on other posts suggested svm instead of softmax in the final layer </p>",
      "votes": null,
      "replies": [
        {
          "id": 1089042,
          "author_name": "dldmw579",
          "author_url": "",
          "post_date": "11/24/2020 06:49:45",
          "content": "<p>Can you share the link of these papers? Thanks</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1089055,
          "author_name": "venkat555",
          "author_url": "",
          "post_date": "11/24/2020 07:04:57",
          "content": "<p><a href=\"https://www.kaggle.com/c/cassava-leaf-disease-classification/discussion/198146\" target=\"_blank\">https://www.kaggle.com/c/cassava-leaf-disease-classification/discussion/198146</a></p>\n<p>iCassava 2019 Fine-Grained Visual Categorization Challenge<br>\nUsing Transfer Learning for Image-Based Cassava Disease Detection</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1090336,
      "author_name": "mrinath",
      "author_url": "",
      "post_date": "11/25/2020 09:08:41",
      "content": "<p>anyone applied label smoothing??</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1144187,
      "author_name": "sayedathar11",
      "author_url": "",
      "post_date": "01/08/2021 09:48:31",
      "content": "<p><a href=\"https://www.kaggle.com/khyeh0719\" target=\"_blank\">@khyeh0719</a> I know what 5 folds and oof means but what does this means save oof predictions for point 7 can you please explain .</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1084768": "1.  Try with different pretrained efficientnet/se-resnext models\n2. Try with different image sizes\n3. Try with different augmentations\n4. Finetune lr scheduling \n5. Try a different optimizer: Adabelief (?)\n6. Train 5-fold models and save the oof predictions for point 7.\n7. Blending\\Stacking\n8. Cutmix with soft labels\n\nAnyone has more ideas to share :) ?",
    "1087062": "Nice tips. What about using an external data set! Is there any resource?",
    "1087100": "Good suggestion!",
    "1088942": "Some of the papers suggested on other posts suggested svm instead of softmax in the final layer",
    "1089042": "Can you share the link of these papers? Thanks",
    "1089055": "https://www.kaggle.com/c/cassava-leaf-disease-classification/discussion/198146\n\niCassava 2019 Fine-Grained Visual Categorization Challenge\nUsing Transfer Learning for Image-Based Cassava Disease Detection",
    "1090336": "anyone applied label smoothing??",
    "1125417": "[7 things that did not worked so far](https://www.kaggle.com/c/cassava-leaf-disease-classification/discussion/205491)\nhere the author says, external dataset didn't help him that much.",
    "1144187": "khyeh0719 I know what 5 folds and oof means but what does this means save oof predictions for point 7 can you please explain ."
  },
  "source": "meta"
}