{
  "id": 48759,
  "title": "train set split to categories (files here)",
  "url": "/competitions/data-science-bowl-2018/discussion/48759",
  "author_name": "",
  "post_date": "2018-02-01T16:42:27.979094400Z",
  "votes": 3,
  "comment_count": 6,
  "views": 0,
  "content": "<p>i put here the images and masks (as one pictures) split to categories. every mask named the same as her picture the same... <br>\nthis way u can easily use it for train it by categories to get maybe better results, as suggested here in some discussions \nu can easily use keras \"flow from directory\"</p>\n\n<p>i have here 7 categories and i think if u want u can split to more</p>",
  "messages": [
    {
      "id": "276843",
      "postDate": "02/01/2018 16:42:27",
      "content": "<p>i put here the images and masks (as one pictures) split to categories. every mask named the same as her picture the same... <br>\nthis way u can easily use it for train it by categories to get maybe better results, as suggested here in some discussions \nu can easily use keras \"flow from directory\"</p>\n\n<p>i have here 7 categories and i think if u want u can split to more</p>",
      "rawMarkdown": "i put here the images and masks (as one pictures) split to categories. every mask named the same as her picture the same...  \nthis way u can easily use it for train it by categories to get maybe better results, as suggested here in some discussions \nu can easily use keras \"flow from directory\"\n\ni have here 7 categories and i think if u want u can split to more",
      "votes": null
    },
    {
      "id": "276845",
      "postDate": "02/01/2018 16:49:07",
      "content": "<p>Hi, could you elaborate on how you split the images into the 7 categories? </p>\n\n<blockquote>\n  <p>Ah found the kernel. </p>\n</blockquote>\n\n<p><img src=\"https://www.kaggleusercontent.com/kf/2164702/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..r8kjTRXysHejnco9AT9Jbw.zKFLsWYdTXUztbvCViMAm_liztkUzvHPAeH1m_uzWpaFQNC62br4A-ikBbE8Umbu8nykg8IZwlJlDsLE9lbWXScg3DbWvOZIX_l24lZ0jyv3AuY2GHdvF06AkkufAgLEUa16poG1vRPWZZOFKoZzgw.fpr3fh5yVRat__vzvAmv1g/__results___files/__results___10_0.png\" alt=\"the 7 categories\"></p>",
      "rawMarkdown": "Hi, could you elaborate on how you split the images into the 7 categories? \n\n&gt; Ah found the kernel. \n\n![the 7 categories][1]\n\n\n  [1]: https://www.kaggleusercontent.com/kf/2164702/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..r8kjTRXysHejnco9AT9Jbw.zKFLsWYdTXUztbvCViMAm_liztkUzvHPAeH1m_uzWpaFQNC62br4A-ikBbE8Umbu8nykg8IZwlJlDsLE9lbWXScg3DbWvOZIX_l24lZ0jyv3AuY2GHdvF06AkkufAgLEUa16poG1vRPWZZOFKoZzgw.fpr3fh5yVRat__vzvAmv1g/__results___files/__results___10_0.png",
      "votes": null
    },
    {
      "id": "276847",
      "postDate": "02/01/2018 17:03:44",
      "content": "<blockquote>\n  <p>Hi, could you elaborate on how you split the images into the 7\n  categories?</p>\n</blockquote>\n\n<p>its a big dilema because i think there are many aspects to that. like in the white purple do u want to split by diff size.\nthere are two types there... </p>\n\n<p>i splited to seven only the train set. i didnt split the test. basicly because i dont train on it... </p>",
      "rawMarkdown": "&gt; Hi, could you elaborate on how you split the images into the 7\n&gt; categories?\n\nits a big dilema because i think there are many aspects to that. like in the white purple do u want to split by diff size.\nthere are two types there... \n\ni splited to seven only the train set. i didnt split the test. basicly because i dont train on it...",
      "votes": null
    },
    {
      "id": "277092",
      "postDate": "02/02/2018 05:27:05",
      "content": "<p>Could you please provide the code that is used to derive this separation? I can see kernels that do 3-subclass seperation, but I couldn't find the 7-class ones. </p>",
      "rawMarkdown": "Could you please provide the code that is used to derive this separation? I can see kernels that do 3-subclass seperation, but I couldn't find the 7-class ones.",
      "votes": null
    },
    {
      "id": "277159",
      "postDate": "02/02/2018 09:48:04",
      "content": "<p>actually i saved all the images and the labels to one libary and i did it by hand. i looked at the pictures and chose what i think my eye seperate</p>",
      "rawMarkdown": "actually i saved all the images and the labels to one libary and i did it by hand. i looked at the pictures and chose what i think my eye seperate",
      "votes": null
    },
    {
      "id": "277167",
      "postDate": "02/02/2018 10:29:35",
      "content": "<p>I see. Is there any way to get the mapping from names of the images you provided to the original image names?</p>",
      "rawMarkdown": "I see. Is there any way to get the mapping from names of the images you provided to the original image names?",
      "votes": null
    },
    {
      "id": "277600",
      "postDate": "02/03/2018 17:13:14",
      "content": "<p>its the order of the X_train of \"keras unet startet\" kernel. u can take the list of files drom there and map it from 1 to 670</p>",
      "rawMarkdown": "its the order of the X_train of \"keras unet startet\" kernel. u can take the list of files drom there and map it from 1 to 670",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 276845,
      "author_name": "etheleon",
      "author_url": "",
      "post_date": "02/01/2018 16:49:07",
      "content": "<p>Hi, could you elaborate on how you split the images into the 7 categories? </p>\n\n<blockquote>\n  <p>Ah found the kernel. </p>\n</blockquote>\n\n<p><img src=\"https://www.kaggleusercontent.com/kf/2164702/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..r8kjTRXysHejnco9AT9Jbw.zKFLsWYdTXUztbvCViMAm_liztkUzvHPAeH1m_uzWpaFQNC62br4A-ikBbE8Umbu8nykg8IZwlJlDsLE9lbWXScg3DbWvOZIX_l24lZ0jyv3AuY2GHdvF06AkkufAgLEUa16poG1vRPWZZOFKoZzgw.fpr3fh5yVRat__vzvAmv1g/__results___files/__results___10_0.png\" alt=\"the 7 categories\"></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 276847,
      "author_name": "aviwolfson",
      "author_url": "",
      "post_date": "02/01/2018 17:03:44",
      "content": "<blockquote>\n  <p>Hi, could you elaborate on how you split the images into the 7\n  categories?</p>\n</blockquote>\n\n<p>its a big dilema because i think there are many aspects to that. like in the white purple do u want to split by diff size.\nthere are two types there... </p>\n\n<p>i splited to seven only the train set. i didnt split the test. basicly because i dont train on it... </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 277092,
      "author_name": "ukrdailo",
      "author_url": "",
      "post_date": "02/02/2018 05:27:05",
      "content": "<p>Could you please provide the code that is used to derive this separation? I can see kernels that do 3-subclass seperation, but I couldn't find the 7-class ones. </p>",
      "votes": null,
      "replies": [
        {
          "id": 277159,
          "author_name": "aviwolfson",
          "author_url": "",
          "post_date": "02/02/2018 09:48:04",
          "content": "<p>actually i saved all the images and the labels to one libary and i did it by hand. i looked at the pictures and chose what i think my eye seperate</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 277167,
          "author_name": "ukrdailo",
          "author_url": "",
          "post_date": "02/02/2018 10:29:35",
          "content": "<p>I see. Is there any way to get the mapping from names of the images you provided to the original image names?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 277600,
          "author_name": "aviwolfson",
          "author_url": "",
          "post_date": "02/03/2018 17:13:14",
          "content": "<p>its the order of the X_train of \"keras unet startet\" kernel. u can take the list of files drom there and map it from 1 to 670</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "276843": "i put here the images and masks (as one pictures) split to categories. every mask named the same as her picture the same...  \nthis way u can easily use it for train it by categories to get maybe better results, as suggested here in some discussions \nu can easily use keras \"flow from directory\"\n\ni have here 7 categories and i think if u want u can split to more",
    "276845": "Hi, could you elaborate on how you split the images into the 7 categories? \n\n&gt; Ah found the kernel. \n\n![the 7 categories][1]\n\n\n  [1]: https://www.kaggleusercontent.com/kf/2164702/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..r8kjTRXysHejnco9AT9Jbw.zKFLsWYdTXUztbvCViMAm_liztkUzvHPAeH1m_uzWpaFQNC62br4A-ikBbE8Umbu8nykg8IZwlJlDsLE9lbWXScg3DbWvOZIX_l24lZ0jyv3AuY2GHdvF06AkkufAgLEUa16poG1vRPWZZOFKoZzgw.fpr3fh5yVRat__vzvAmv1g/__results___files/__results___10_0.png",
    "276847": "&gt; Hi, could you elaborate on how you split the images into the 7\n&gt; categories?\n\nits a big dilema because i think there are many aspects to that. like in the white purple do u want to split by diff size.\nthere are two types there... \n\ni splited to seven only the train set. i didnt split the test. basicly because i dont train on it...",
    "277092": "Could you please provide the code that is used to derive this separation? I can see kernels that do 3-subclass seperation, but I couldn't find the 7-class ones.",
    "277159": "actually i saved all the images and the labels to one libary and i did it by hand. i looked at the pictures and chose what i think my eye seperate",
    "277167": "I see. Is there any way to get the mapping from names of the images you provided to the original image names?",
    "277600": "its the order of the X_train of \"keras unet startet\" kernel. u can take the list of files drom there and map it from 1 to 670"
  },
  "source": "meta"
}