{
  "id": 95332,
  "title": "Can anyone explain how to generate prediction string?",
  "url": "/competitions/open-images-2019-object-detection/discussion/95332",
  "author_name": "",
  "post_date": "2019-06-11T14:29:29.213784700Z",
  "votes": 8,
  "comment_count": 9,
  "views": 0,
  "content": "",
  "messages": [
    {
      "id": "550351",
      "postDate": "06/11/2019 14:29:29",
      "content": "",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "550355",
      "postDate": "06/11/2019 14:33:04",
      "content": "<p>I mean if our image is '/m/01s01s' it has 2 object detected with (x1top,y1top),(x1bottom,y1bottom) ,(x2top,y2top),(x2bottom,y2bottom)  with 0.91 and 0.92 probability then how to construct prediction string?</p>\n\n<p>How below string is constructed and what does it's each part means?\n/m/05s2s 0.9 0.46 0.08 0.93 0.5 /m/0c9ph5 0.5 0.25 0.6 0.6 0.9</p>\n\n<p>Thanks!!!! :):):)</p>",
      "rawMarkdown": "I mean if our image is '/m/01s01s' it has 2 object detected with (x1top,y1top),(x1bottom,y1bottom) ,(x2top,y2top),(x2bottom,y2bottom)  with 0.91 and 0.92 probability then how to construct prediction string?\n\nHow below string is constructed and what does it's each part means?\n/m/05s2s 0.9 0.46 0.08 0.93 0.5 /m/0c9ph5 0.5 0.25 0.6 0.6 0.9\n\nThanks!!!! :):):)",
      "votes": null
    },
    {
      "id": "550361",
      "postDate": "06/11/2019 14:38:57",
      "content": "<p>is LabelName,Confidence,XMin,YMin,XMax,YMax  ？?？</p>",
      "rawMarkdown": "is LabelName,Confidence,XMin,YMin,XMax,YMax  ？?？",
      "votes": null
    },
    {
      "id": "550689",
      "postDate": "06/11/2019 23:46:43",
      "content": "<p>/m/01s01s in your example is the object detected in the image, not the image itself, after which you give the models confidence that the thing is actually from that class.  </p>\n\n<p>Each of these object detections is specified with a bounding box given by the coordinates of its 4 corners. </p>",
      "rawMarkdown": "/m/01s01s in your example is the object detected in the image, not the image itself, after which you give the models confidence that the thing is actually from that class.  \n\nEach of these object detections is specified with a bounding box given by the coordinates of its 4 corners.",
      "votes": null
    },
    {
      "id": "550879",
      "postDate": "06/12/2019 05:40:24",
      "content": "<p>Thank you so much for your help</p>",
      "rawMarkdown": "Thank you so much for your help",
      "votes": null
    },
    {
      "id": "550880",
      "postDate": "06/12/2019 05:40:41",
      "content": "<p>Thank you so much for your help</p>",
      "rawMarkdown": "Thank you so much for your help",
      "votes": null
    },
    {
      "id": "555442",
      "postDate": "06/18/2019 22:13:28",
      "content": "<p>So, \"/m/01s01s\" is the label for the class that is detected. Right?</p>",
      "rawMarkdown": "So, \"/m/01s01s\" is the label for the class that is detected. Right?",
      "votes": null
    },
    {
      "id": "555591",
      "postDate": "06/19/2019 05:55:52",
      "content": "<p>i hope so,\nand it sounds like the same.</p>",
      "rawMarkdown": "i hope so,\nand it sounds like the same.",
      "votes": null
    },
    {
      "id": "571114",
      "postDate": "07/09/2019 07:18:09",
      "content": "<p>If you can successfully draw boxes, then you can just create your PredictionString like: </p>\n\n<p>/m/0dzct 0.524965 0.004842 0.002574 0.869498 1.000000 /m/0dzct 0.213452 0.000000 0.113544 0.392445 0.992020 ...</p>\n\n<p>which is just: </p>\n\n<p>class_label1 probs x1 y2 x2 y2 class_label2 probs x1 y2 x2 y2 ....</p>",
      "rawMarkdown": "If you can successfully draw boxes, then you can just create your PredictionString like: \n\n/m/0dzct 0.524965 0.004842 0.002574 0.869498 1.000000 /m/0dzct 0.213452 0.000000 0.113544 0.392445 0.992020 ...\n\nwhich is just: \n\nclass_label1 probs x1 y2 x2 y2 class_label2 probs x1 y2 x2 y2 ....",
      "votes": null
    },
    {
      "id": "849987",
      "postDate": "05/16/2020 08:31:57",
      "content": "<p>hi\nCan you tell me any one..\nhow to write the code for one image?</p>",
      "rawMarkdown": "hi\nCan you tell me any one..\nhow to write the code for one image?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 550355,
      "author_name": "siddhrath",
      "author_url": "",
      "post_date": "06/11/2019 14:33:04",
      "content": "<p>I mean if our image is '/m/01s01s' it has 2 object detected with (x1top,y1top),(x1bottom,y1bottom) ,(x2top,y2top),(x2bottom,y2bottom)  with 0.91 and 0.92 probability then how to construct prediction string?</p>\n\n<p>How below string is constructed and what does it's each part means?\n/m/05s2s 0.9 0.46 0.08 0.93 0.5 /m/0c9ph5 0.5 0.25 0.6 0.6 0.9</p>\n\n<p>Thanks!!!! :):):)</p>",
      "votes": null,
      "replies": [
        {
          "id": 550689,
          "author_name": "interneuron",
          "author_url": "",
          "post_date": "06/11/2019 23:46:43",
          "content": "<p>/m/01s01s in your example is the object detected in the image, not the image itself, after which you give the models confidence that the thing is actually from that class.  </p>\n\n<p>Each of these object detections is specified with a bounding box given by the coordinates of its 4 corners. </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 550880,
          "author_name": "siddhrath",
          "author_url": "",
          "post_date": "06/12/2019 05:40:41",
          "content": "<p>Thank you so much for your help</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 555442,
          "author_name": "manishrai",
          "author_url": "",
          "post_date": "06/18/2019 22:13:28",
          "content": "<p>So, \"/m/01s01s\" is the label for the class that is detected. Right?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 555591,
          "author_name": "siddhrath",
          "author_url": "",
          "post_date": "06/19/2019 05:55:52",
          "content": "<p>i hope so,\nand it sounds like the same.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 550361,
      "author_name": "withli",
      "author_url": "",
      "post_date": "06/11/2019 14:38:57",
      "content": "<p>is LabelName,Confidence,XMin,YMin,XMax,YMax  ？?？</p>",
      "votes": null,
      "replies": [
        {
          "id": 550879,
          "author_name": "siddhrath",
          "author_url": "",
          "post_date": "06/12/2019 05:40:24",
          "content": "<p>Thank you so much for your help</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 571114,
      "author_name": "yw6916",
      "author_url": "",
      "post_date": "07/09/2019 07:18:09",
      "content": "<p>If you can successfully draw boxes, then you can just create your PredictionString like: </p>\n\n<p>/m/0dzct 0.524965 0.004842 0.002574 0.869498 1.000000 /m/0dzct 0.213452 0.000000 0.113544 0.392445 0.992020 ...</p>\n\n<p>which is just: </p>\n\n<p>class_label1 probs x1 y2 x2 y2 class_label2 probs x1 y2 x2 y2 ....</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 849987,
      "author_name": "sandhya9173",
      "author_url": "",
      "post_date": "05/16/2020 08:31:57",
      "content": "<p>hi\nCan you tell me any one..\nhow to write the code for one image?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "550351": "",
    "550355": "I mean if our image is '/m/01s01s' it has 2 object detected with (x1top,y1top),(x1bottom,y1bottom) ,(x2top,y2top),(x2bottom,y2bottom)  with 0.91 and 0.92 probability then how to construct prediction string?\n\nHow below string is constructed and what does it's each part means?\n/m/05s2s 0.9 0.46 0.08 0.93 0.5 /m/0c9ph5 0.5 0.25 0.6 0.6 0.9\n\nThanks!!!! :):):)",
    "550361": "is LabelName,Confidence,XMin,YMin,XMax,YMax  ？?？",
    "550689": "/m/01s01s in your example is the object detected in the image, not the image itself, after which you give the models confidence that the thing is actually from that class.  \n\nEach of these object detections is specified with a bounding box given by the coordinates of its 4 corners.",
    "550879": "Thank you so much for your help",
    "550880": "Thank you so much for your help",
    "555442": "So, \"/m/01s01s\" is the label for the class that is detected. Right?",
    "555591": "i hope so,\nand it sounds like the same.",
    "571114": "If you can successfully draw boxes, then you can just create your PredictionString like: \n\n/m/0dzct 0.524965 0.004842 0.002574 0.869498 1.000000 /m/0dzct 0.213452 0.000000 0.113544 0.392445 0.992020 ...\n\nwhich is just: \n\nclass_label1 probs x1 y2 x2 y2 class_label2 probs x1 y2 x2 y2 ....",
    "849987": "hi\nCan you tell me any one..\nhow to write the code for one image?"
  },
  "source": "meta"
}