{
  "id": 130678,
  "title": "How to create resized image dataset and save as kernels output for faster training? ",
  "url": "/competitions/bengaliai-cv19/discussion/130678",
  "author_name": "",
  "post_date": "2020-02-15T17:38:15.386305400Z",
  "votes": 25,
  "comment_count": 10,
  "views": 0,
  "content": "<p>i have created 128x128x3 png image dataset for this competition,using my kernel you can create 64x64x3 png image dataset as well and train your deep models much longer.\nfor saving time and money, here is : \n1 . Dataset : <a href=\"https://www.kaggle.com/mobassir/bengaliai128\"><strong>bengaliai128</strong></a></p>\n\n<ol>\n<li>kernel for generating such datasets : <a href=\"https://www.kaggle.com/mobassir/resized-rgb-images-for-faster-training\"><strong>Resized RGB images for faster training</strong></a></li>\n</ol>\n\n<p>asking for help : \"if anyone has feather dataset of 128x128x1 or 256x256x1 then will you please share them with us? 🙏 \" thank you a lot in advance. 😊 👇 </p>",
  "messages": [
    {
      "id": "746894",
      "postDate": "02/15/2020 17:38:15",
      "content": "<p>i have created 128x128x3 png image dataset for this competition,using my kernel you can create 64x64x3 png image dataset as well and train your deep models much longer.\nfor saving time and money, here is : \n1 . Dataset : <a href=\"https://www.kaggle.com/mobassir/bengaliai128\"><strong>bengaliai128</strong></a></p>\n\n<ol>\n<li>kernel for generating such datasets : <a href=\"https://www.kaggle.com/mobassir/resized-rgb-images-for-faster-training\"><strong>Resized RGB images for faster training</strong></a></li>\n</ol>\n\n<p>asking for help : \"if anyone has feather dataset of 128x128x1 or 256x256x1 then will you please share them with us? 🙏 \" thank you a lot in advance. 😊 👇 </p>",
      "rawMarkdown": "i have created 128x128x3 png image dataset for this competition,using my kernel you can create 64x64x3 png image dataset as well and train your deep models much longer.\nfor saving time and money, here is : \n1 . Dataset : [**bengaliai128**](https://www.kaggle.com/mobassir/bengaliai128)\n\n2. kernel for generating such datasets : [**Resized RGB images for faster training**](https://www.kaggle.com/mobassir/resized-rgb-images-for-faster-training)\n\nasking for help : \"if anyone has feather dataset of 128x128x1 or 256x256x1 then will you please share them with us? 🙏 \" thank you a lot in advance. 😊 👇",
      "votes": null
    },
    {
      "id": "747153",
      "postDate": "02/16/2020 03:26:09",
      "content": "<p>Thanks for Sharing!! <a href=\"/mobassir\">@mobassir</a> </p>",
      "rawMarkdown": "Thanks for Sharing!! @mobassir",
      "votes": null
    },
    {
      "id": "747231",
      "postDate": "02/16/2020 06:35:19",
      "content": "<p>Thanks for Sharing!!</p>",
      "rawMarkdown": "Thanks for Sharing!!",
      "votes": null
    },
    {
      "id": "747979",
      "postDate": "02/17/2020 04:10:22",
      "content": "<p>How much time does this save vs just concatenating the single channel images?</p>",
      "rawMarkdown": "How much time does this save vs just concatenating the single channel images?",
      "votes": null
    },
    {
      "id": "748342",
      "postDate": "02/17/2020 12:12:00",
      "content": "<p>concatenating single channel will be faster but we can't apply transfer learning there but still it is a good idea to use single channel image <a href=\"/greatgamedota\">@greatgamedota</a> </p>",
      "rawMarkdown": "concatenating single channel will be faster but we can't apply transfer learning there but still it is a good idea to use single channel image @greatgamedota",
      "votes": null
    },
    {
      "id": "748667",
      "postDate": "02/17/2020 19:58:55",
      "content": "<p>Thanks for sharing it brother!</p>",
      "rawMarkdown": "Thanks for sharing it brother!",
      "votes": null
    },
    {
      "id": "748679",
      "postDate": "02/17/2020 20:33:45",
      "content": "<p>my pleasure</p>",
      "rawMarkdown": "my pleasure",
      "votes": null
    },
    {
      "id": "749913",
      "postDate": "02/19/2020 02:09:37",
      "content": "<p>If anyone want 128x128 dataset with single channel here is the link to it. <a href=\"https://www.kaggle.com/ratan123/128x128x1-bengali-feather-data\">https://www.kaggle.com/ratan123/128x128x1-bengali-feather-data</a> \nHave a great day 😄 </p>",
      "rawMarkdown": "If anyone want 128x128 dataset with single channel here is the link to it. https://www.kaggle.com/ratan123/128x128x1-bengali-feather-data \nHave a great day 😄",
      "votes": null
    },
    {
      "id": "755764",
      "postDate": "02/25/2020 05:27:38",
      "content": "<p>Thanks for Sharing!!</p>",
      "rawMarkdown": "Thanks for Sharing!!",
      "votes": null
    },
    {
      "id": "755790",
      "postDate": "02/25/2020 06:03:09",
      "content": "<p>Thanks Mobassir</p>",
      "rawMarkdown": "Thanks Mobassir",
      "votes": null
    },
    {
      "id": "758787",
      "postDate": "02/28/2020 06:39:10",
      "content": "<p>great work! thanks <a href=\"/mobassir\">@mobassir</a> !</p>",
      "rawMarkdown": "great work! thanks @mobassir !",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 747153,
      "author_name": "veeralakrishna",
      "author_url": "",
      "post_date": "02/16/2020 03:26:09",
      "content": "<p>Thanks for Sharing!! <a href=\"/mobassir\">@mobassir</a> </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 747231,
      "author_name": "aravindhkv123",
      "author_url": "",
      "post_date": "02/16/2020 06:35:19",
      "content": "<p>Thanks for Sharing!!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 747979,
      "author_name": "greatgamedota",
      "author_url": "",
      "post_date": "02/17/2020 04:10:22",
      "content": "<p>How much time does this save vs just concatenating the single channel images?</p>",
      "votes": null,
      "replies": [
        {
          "id": 748342,
          "author_name": "mobassir",
          "author_url": "",
          "post_date": "02/17/2020 12:12:00",
          "content": "<p>concatenating single channel will be faster but we can't apply transfer learning there but still it is a good idea to use single channel image <a href=\"/greatgamedota\">@greatgamedota</a> </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 748667,
      "author_name": "sumitm004",
      "author_url": "",
      "post_date": "02/17/2020 19:58:55",
      "content": "<p>Thanks for sharing it brother!</p>",
      "votes": null,
      "replies": [
        {
          "id": 748679,
          "author_name": "mobassir",
          "author_url": "",
          "post_date": "02/17/2020 20:33:45",
          "content": "<p>my pleasure</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 749913,
      "author_name": "ratan123",
      "author_url": "",
      "post_date": "02/19/2020 02:09:37",
      "content": "<p>If anyone want 128x128 dataset with single channel here is the link to it. <a href=\"https://www.kaggle.com/ratan123/128x128x1-bengali-feather-data\">https://www.kaggle.com/ratan123/128x128x1-bengali-feather-data</a> \nHave a great day 😄 </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 755764,
      "author_name": "zhangliao",
      "author_url": "",
      "post_date": "02/25/2020 05:27:38",
      "content": "<p>Thanks for Sharing!!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 755790,
      "author_name": "cdeotte",
      "author_url": "",
      "post_date": "02/25/2020 06:03:09",
      "content": "<p>Thanks Mobassir</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 758787,
      "author_name": "aleksandradeis",
      "author_url": "",
      "post_date": "02/28/2020 06:39:10",
      "content": "<p>great work! thanks <a href=\"/mobassir\">@mobassir</a> !</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "746894": "i have created 128x128x3 png image dataset for this competition,using my kernel you can create 64x64x3 png image dataset as well and train your deep models much longer.\nfor saving time and money, here is : \n1 . Dataset : [**bengaliai128**](https://www.kaggle.com/mobassir/bengaliai128)\n\n2. kernel for generating such datasets : [**Resized RGB images for faster training**](https://www.kaggle.com/mobassir/resized-rgb-images-for-faster-training)\n\nasking for help : \"if anyone has feather dataset of 128x128x1 or 256x256x1 then will you please share them with us? 🙏 \" thank you a lot in advance. 😊 👇",
    "747153": "Thanks for Sharing!! @mobassir",
    "747231": "Thanks for Sharing!!",
    "747979": "How much time does this save vs just concatenating the single channel images?",
    "748342": "concatenating single channel will be faster but we can't apply transfer learning there but still it is a good idea to use single channel image @greatgamedota",
    "748667": "Thanks for sharing it brother!",
    "748679": "my pleasure",
    "749913": "If anyone want 128x128 dataset with single channel here is the link to it. https://www.kaggle.com/ratan123/128x128x1-bengali-feather-data \nHave a great day 😄",
    "755764": "Thanks for Sharing!!",
    "755790": "Thanks Mobassir",
    "758787": "great work! thanks @mobassir !"
  },
  "source": "meta"
}