{
  "id": 245060,
  "title": "\"Snap to Grid\" is valid for this competition.",
  "url": "/competitions/google-smartphone-decimeter-challenge/discussion/245060",
  "author_name": "",
  "post_date": "2021-06-09T13:47:21.782277300Z",
  "votes": 38,
  "comment_count": 11,
  "views": 0,
  "content": "<p>I tried post processing called by \"<a href=\"https://www.kaggle.com/robikscube/indoor-navigation-snap-to-grid-post-processing\" target=\"_blank\">Snap to Grid</a>\", which was commonly used in <a href=\"https://www.kaggle.com/c/indoor-location-navigation/overview\" target=\"_blank\">Indoor Location &amp; Navigation</a>.</p>\n<p>It is effective for me and LB score is improve 5.630 -&gt; 5.429.<br>\nPlease refer <a href=\"https://www.kaggle.com/robikscube/indoor-navigation-snap-to-grid-post-processing\" target=\"_blank\">original notebook</a> for more details about Snap to grid.</p>\n<p>And I also shared <a href=\"https://www.kaggle.com/kuto0633/road-detection-and-creating-grid-points\" target=\"_blank\">the notebook</a> to detect the roads and  to generate grid points for \"Snap to Grid\".</p>\n<p>I hope it helps. Thanks!</p>",
  "messages": [
    {
      "id": "1342483",
      "postDate": "06/09/2021 13:47:21",
      "content": "<p>I tried post processing called by \"<a href=\"https://www.kaggle.com/robikscube/indoor-navigation-snap-to-grid-post-processing\" target=\"_blank\">Snap to Grid</a>\", which was commonly used in <a href=\"https://www.kaggle.com/c/indoor-location-navigation/overview\" target=\"_blank\">Indoor Location &amp; Navigation</a>.</p>\n<p>It is effective for me and LB score is improve 5.630 -&gt; 5.429.<br>\nPlease refer <a href=\"https://www.kaggle.com/robikscube/indoor-navigation-snap-to-grid-post-processing\" target=\"_blank\">original notebook</a> for more details about Snap to grid.</p>\n<p>And I also shared <a href=\"https://www.kaggle.com/kuto0633/road-detection-and-creating-grid-points\" target=\"_blank\">the notebook</a> to detect the roads and  to generate grid points for \"Snap to Grid\".</p>\n<p>I hope it helps. Thanks!</p>",
      "rawMarkdown": "I tried post processing called by \"[Snap to Grid](https://www.kaggle.com/robikscube/indoor-navigation-snap-to-grid-post-processing)\", which was commonly used in [Indoor Location & Navigation](https://www.kaggle.com/c/indoor-location-navigation/overview).\n\nIt is effective for me and LB score is improve 5.630 -> 5.429.\nPlease refer [original notebook](https://www.kaggle.com/robikscube/indoor-navigation-snap-to-grid-post-processing) for more details about Snap to grid.\n\nAnd I also shared [the notebook](https://www.kaggle.com/kuto0633/road-detection-and-creating-grid-points) to detect the roads and  to generate grid points for \"Snap to Grid\".\n\nI hope it helps. Thanks!",
      "votes": null
    },
    {
      "id": "1342500",
      "postDate": "06/09/2021 14:04:32",
      "content": "<p>Thank you for a great topic and notebook!!<br>\nOur team has been interpolating ground_truth and doing snap to grid, but we were not aware of a way to use the geography library. It looks like your way is more accurate!</p>",
      "rawMarkdown": "Thank you for a great topic and notebook!!\nOur team has been interpolating ground_truth and doing snap to grid, but we were not aware of a way to use the geography library. It looks like your way is more accurate!",
      "votes": null
    },
    {
      "id": "1342514",
      "postDate": "06/09/2021 14:17:28",
      "content": "<p>It was very helpful. Thank you very much.</p>",
      "rawMarkdown": "It was very helpful. Thank you very much.",
      "votes": null
    },
    {
      "id": "1342566",
      "postDate": "06/09/2021 14:52:13",
      "content": "<blockquote>\n  <p>Our team has been interpolating ground_truth and doing snap to grid,</p>\n</blockquote>\n<p>Thanks for sharing! I also tried Snap to Grid by using ground_truth.csv.<br>\nCurrently, the score is better on  using grond truth, and I am in the process of trial and error, for apply generated road grids.</p>",
      "rawMarkdown": "> Our team has been interpolating ground_truth and doing snap to grid,\n\nThanks for sharing! I also tried Snap to Grid by using ground_truth.csv.\nCurrently, the score is better on  using grond truth, and I am in the process of trial and error, for apply generated road grids.",
      "votes": null
    },
    {
      "id": "1342824",
      "postDate": "06/09/2021 18:37:11",
      "content": "<p>Nice, mine didn't bring for improvement</p>",
      "rawMarkdown": "Nice, mine didn't bring for improvement",
      "votes": null
    },
    {
      "id": "1342901",
      "postDate": "06/09/2021 20:14:12",
      "content": "<p>Thank you for sharing your great notebook.<br>\nI also tried Snap to Grid, but it failed because it didn't detect the road.Your idea is very helpful for me.</p>",
      "rawMarkdown": "Thank you for sharing your great notebook.\nI also tried Snap to Grid, but it failed because it didn't detect the road.Your idea is very helpful for me.",
      "votes": null
    },
    {
      "id": "1342919",
      "postDate": "06/09/2021 20:25:02",
      "content": "<p>Thanks. I’m grad it helps for you!</p>",
      "rawMarkdown": "Thanks. I’m grad it helps for you!",
      "votes": null
    },
    {
      "id": "1344181",
      "postDate": "06/10/2021 17:09:16",
      "content": "<p>Thank you for your great sharing!!</p>\n<p>I think there are useful case and difficult case.</p>\n<p>Useful case: The road is narrow or base point is not on the correct road or the road is not contained train ground truth. <br>\nDifficult case: The road is wide (multiple car lines) or base point is on the correct road</p>\n<p>In difficult case, using ground truth may be better than map information.</p>\n<p>However I have no idea how to combine these logic.</p>",
      "rawMarkdown": "Thank you for your great sharing!!\n\nI think there are useful case and difficult case.\n\nUseful case: The road is narrow or base point is not on the correct road or the road is not contained train ground truth. \nDifficult case: The road is wide (multiple car lines) or base point is on the correct road\n\nIn difficult case, using ground truth may be better than map information.\n\nHowever I have no idea how to combine these logic.",
      "votes": null
    },
    {
      "id": "1344528",
      "postDate": "06/11/2021 01:50:26",
      "content": "<p>Thanks for your comment!<br>\nI agree,  there are some difficult case to apply it.</p>\n<p>In indoor Location &amp; Navigation competition, there were some technics to improve the effect of Snap to Grid. So I share.</p>\n<ul>\n<li>using both ground_truth grids and generated grids as grids</li>\n<li>setting threshold for snapping distance(min/max) </li>\n<li>changing the generated grids density</li>\n<li>selecting correct grids by discrete optimization</li>\n</ul>\n<p>I don’t make sure, but these may be useful.</p>",
      "rawMarkdown": "Thanks for your comment!\nI agree,  there are some difficult case to apply it.\n\nIn indoor Location & Navigation competition, there were some technics to improve the effect of Snap to Grid. So I share.\n- using both ground_truth grids and generated grids as grids\n- setting threshold for snapping distance(min/max) \n- changing the generated grids density\n- selecting correct grids by discrete optimization\n\nI don’t make sure, but these may be useful.",
      "votes": null
    },
    {
      "id": "1344623",
      "postDate": "06/11/2021 04:04:40",
      "content": "<p>Thank you for sharing indoor competition knowledge!!</p>\n<p>In mapping strategy, we have many choices and a room to improve :)</p>",
      "rawMarkdown": "Thank you for sharing indoor competition knowledge!!\n\nIn mapping strategy, we have many choices and a room to improve :)",
      "votes": null
    },
    {
      "id": "1345494",
      "postDate": "06/11/2021 16:08:38",
      "content": "<p>Thanks for sharing, very useful!</p>",
      "rawMarkdown": "Thanks for sharing, very useful!",
      "votes": null
    },
    {
      "id": "1361493",
      "postDate": "06/22/2021 21:38:34",
      "content": "<p>Very exhaustive NB. Congrats !!</p>",
      "rawMarkdown": "Very exhaustive NB. Congrats !!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1342500,
      "author_name": "ryotayoshinobu",
      "author_url": "",
      "post_date": "06/09/2021 14:04:32",
      "content": "<p>Thank you for a great topic and notebook!!<br>\nOur team has been interpolating ground_truth and doing snap to grid, but we were not aware of a way to use the geography library. It looks like your way is more accurate!</p>",
      "votes": null,
      "replies": [
        {
          "id": 1342566,
          "author_name": "kuto0633",
          "author_url": "",
          "post_date": "06/09/2021 14:52:13",
          "content": "<blockquote>\n  <p>Our team has been interpolating ground_truth and doing snap to grid,</p>\n</blockquote>\n<p>Thanks for sharing! I also tried Snap to Grid by using ground_truth.csv.<br>\nCurrently, the score is better on  using grond truth, and I am in the process of trial and error, for apply generated road grids.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1342514,
      "author_name": "zakopur0",
      "author_url": "",
      "post_date": "06/09/2021 14:17:28",
      "content": "<p>It was very helpful. Thank you very much.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1342824,
      "author_name": "ulrich07",
      "author_url": "",
      "post_date": "06/09/2021 18:37:11",
      "content": "<p>Nice, mine didn't bring for improvement</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1342901,
      "author_name": "dehokanta",
      "author_url": "",
      "post_date": "06/09/2021 20:14:12",
      "content": "<p>Thank you for sharing your great notebook.<br>\nI also tried Snap to Grid, but it failed because it didn't detect the road.Your idea is very helpful for me.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1342919,
          "author_name": "kuto0633",
          "author_url": "",
          "post_date": "06/09/2021 20:25:02",
          "content": "<p>Thanks. I’m grad it helps for you!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1344181,
      "author_name": "go5kuramubon",
      "author_url": "",
      "post_date": "06/10/2021 17:09:16",
      "content": "<p>Thank you for your great sharing!!</p>\n<p>I think there are useful case and difficult case.</p>\n<p>Useful case: The road is narrow or base point is not on the correct road or the road is not contained train ground truth. <br>\nDifficult case: The road is wide (multiple car lines) or base point is on the correct road</p>\n<p>In difficult case, using ground truth may be better than map information.</p>\n<p>However I have no idea how to combine these logic.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1344528,
          "author_name": "kuto0633",
          "author_url": "",
          "post_date": "06/11/2021 01:50:26",
          "content": "<p>Thanks for your comment!<br>\nI agree,  there are some difficult case to apply it.</p>\n<p>In indoor Location &amp; Navigation competition, there were some technics to improve the effect of Snap to Grid. So I share.</p>\n<ul>\n<li>using both ground_truth grids and generated grids as grids</li>\n<li>setting threshold for snapping distance(min/max) </li>\n<li>changing the generated grids density</li>\n<li>selecting correct grids by discrete optimization</li>\n</ul>\n<p>I don’t make sure, but these may be useful.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1344623,
          "author_name": "go5kuramubon",
          "author_url": "",
          "post_date": "06/11/2021 04:04:40",
          "content": "<p>Thank you for sharing indoor competition knowledge!!</p>\n<p>In mapping strategy, we have many choices and a room to improve :)</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1345494,
      "author_name": "saurabhbagchi",
      "author_url": "",
      "post_date": "06/11/2021 16:08:38",
      "content": "<p>Thanks for sharing, very useful!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1361493,
      "author_name": "paulronald",
      "author_url": "",
      "post_date": "06/22/2021 21:38:34",
      "content": "<p>Very exhaustive NB. Congrats !!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1342483": "I tried post processing called by \"[Snap to Grid](https://www.kaggle.com/robikscube/indoor-navigation-snap-to-grid-post-processing)\", which was commonly used in [Indoor Location & Navigation](https://www.kaggle.com/c/indoor-location-navigation/overview).\n\nIt is effective for me and LB score is improve 5.630 -> 5.429.\nPlease refer [original notebook](https://www.kaggle.com/robikscube/indoor-navigation-snap-to-grid-post-processing) for more details about Snap to grid.\n\nAnd I also shared [the notebook](https://www.kaggle.com/kuto0633/road-detection-and-creating-grid-points) to detect the roads and  to generate grid points for \"Snap to Grid\".\n\nI hope it helps. Thanks!",
    "1342500": "Thank you for a great topic and notebook!!\nOur team has been interpolating ground_truth and doing snap to grid, but we were not aware of a way to use the geography library. It looks like your way is more accurate!",
    "1342514": "It was very helpful. Thank you very much.",
    "1342566": "> Our team has been interpolating ground_truth and doing snap to grid,\n\nThanks for sharing! I also tried Snap to Grid by using ground_truth.csv.\nCurrently, the score is better on  using grond truth, and I am in the process of trial and error, for apply generated road grids.",
    "1342824": "Nice, mine didn't bring for improvement",
    "1342901": "Thank you for sharing your great notebook.\nI also tried Snap to Grid, but it failed because it didn't detect the road.Your idea is very helpful for me.",
    "1342919": "Thanks. I’m grad it helps for you!",
    "1344181": "Thank you for your great sharing!!\n\nI think there are useful case and difficult case.\n\nUseful case: The road is narrow or base point is not on the correct road or the road is not contained train ground truth. \nDifficult case: The road is wide (multiple car lines) or base point is on the correct road\n\nIn difficult case, using ground truth may be better than map information.\n\nHowever I have no idea how to combine these logic.",
    "1344528": "Thanks for your comment!\nI agree,  there are some difficult case to apply it.\n\nIn indoor Location & Navigation competition, there were some technics to improve the effect of Snap to Grid. So I share.\n- using both ground_truth grids and generated grids as grids\n- setting threshold for snapping distance(min/max) \n- changing the generated grids density\n- selecting correct grids by discrete optimization\n\nI don’t make sure, but these may be useful.",
    "1344623": "Thank you for sharing indoor competition knowledge!!\n\nIn mapping strategy, we have many choices and a room to improve :)",
    "1345494": "Thanks for sharing, very useful!",
    "1361493": "Very exhaustive NB. Congrats !!"
  },
  "source": "meta"
}