{
  "id": 256055,
  "title": "How many image layers do you use?",
  "url": "/competitions/seti-breakthrough-listen/discussion/256055",
  "author_name": "Aristarkh_BFG",
  "post_date": "2021-07-30T15:30:35.235000",
  "votes": 2,
  "comment_count": 6,
  "views": 0,
  "content": "<p>I see that some people use all 6 layers. Someone uses 3 layers. My best results are obtained with 3 layers with target. How much do you use?</p>",
  "messages": [
    {
      "id": 1441320,
      "postDate": "2021-08-03T20:07:44.333Z",
      "content": "<p>For me using \"ON\" observation i.e [0,2,4] works better than using all six channels</p>",
      "rawMarkdown": "For me using \"ON\" observation i.e [0,2,4] works better than using all six channels",
      "votes": 1,
      "replies": [
        {
          "id": 1445296,
          "postDate": "2021-08-04T08:22:59.037Z",
          "content": "<p>You are use spatial image?</p>",
          "rawMarkdown": "You are use spatial image?"
        },
        {
          "id": 1445676,
          "postDate": "2021-08-04T09:43:54.143Z",
          "content": "<p>Yes, then I resize to (1,320,320)</p>",
          "rawMarkdown": "Yes, then I resize to (1,320,320)",
          "votes": 1
        }
      ]
    },
    {
      "id": 1405578,
      "postDate": "2021-07-31T04:11:14.397Z",
      "content": "<p>At first time, I use only 3-layers.<br>\nBecause model have more confused if we also used another 3-layers.</p>\n<p>However, In my personal opinion, there are some leakage in the off-channel.<br>\n(I am not sure about this) <br>\nMany values of the images looks like splited in on-channel and off-channel.</p>\n<p>like [0], [2], [4] channel corresponding with [3], [5], [1] channel. And It also can be flipped.<br>\nYou may easily found some images have more noise like Gaussian noise.<br>\nI think that is not guassian noise, just a scar of spliting image in 2 channel.</p>\n<p>There are total 48 combinations for recover original images.<br>\nEven though there are some leakage on off-channel, <br>\nwe can't directly using it without knowing exact information.<br>\nIt just my personal opinion.</p>\n<p>This competition like to find a steganography…</p>",
      "rawMarkdown": "At first time, I use only 3-layers.\nBecause model have more confused if we also used another 3-layers.\n\nHowever, In my personal opinion, there are some leakage in the off-channel.\n(I am not sure about this) \nMany values of the images looks like splited in on-channel and off-channel.\n\nlike [0], [2], [4] channel corresponding with [3], [5], [1] channel. And It also can be flipped.\nYou may easily found some images have more noise like Gaussian noise.\nI think that is not guassian noise, just a scar of spliting image in 2 channel.\n\nThere are total 48 combinations for recover original images.\nEven though there are some leakage on off-channel, \nwe can't directly using it without knowing exact information.\nIt just my personal opinion.\n\nThis competition like to find a steganography...",
      "votes": 2,
      "replies": [
        {
          "id": 1406022,
          "postDate": "2021-07-31T12:17:00.063Z",
          "content": "<p>Sometimes \"off\" signals look similar to the \"on\" signals because most probably man-made noise was present in both signals. </p>\n<p>Not sure if it could be considered as a leakage tho, on the contrary, model may look at the difference between on and off signal and infer whether signal present in the on and not present in the off can be considered \"alien\"</p>",
          "rawMarkdown": "Sometimes \"off\" signals look similar to the \"on\" signals because most probably man-made noise was present in both signals. \n\nNot sure if it could be considered as a leakage tho, on the contrary, model may look at the difference between on and off signal and infer whether signal present in the on and not present in the off can be considered \"alien\"",
          "votes": 1
        }
      ]
    },
    {
      "id": 1405213,
      "postDate": "2021-07-30T15:30:35.237Z",
      "content": "<p>I see that some people use all 6 layers. Someone uses 3 layers. My best results are obtained with 3 layers with target. How much do you use?</p>",
      "rawMarkdown": "I see that some people use all 6 layers. Someone uses 3 layers. My best results are obtained with 3 layers with target. How much do you use?",
      "votes": 2
    },
    {
      "id": 1445959,
      "postDate": "2021-08-04T10:36:20.793Z",
      "content": "<p>I also use only 3 layers (\"ON\").</p>",
      "rawMarkdown": "I also use only 3 layers (\"ON\")."
    }
  ],
  "comments": [
    {
      "id": 1441320,
      "author_name": "Parth Dhameliya",
      "author_url": "",
      "post_date": "2021-08-03T20:07:44.333000",
      "content": "<p>For me using \"ON\" observation i.e [0,2,4] works better than using all six channels</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1445296,
          "author_name": "Aristarkh_BFG",
          "author_url": "",
          "post_date": "2021-08-04T08:22:59.037000",
          "content": "<p>You are use spatial image?</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1445676,
          "author_name": "Parth Dhameliya",
          "author_url": "",
          "post_date": "2021-08-04T09:43:54.143000",
          "content": "<p>Yes, then I resize to (1,320,320)</p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1405578,
      "author_name": "WOOSUNG YOON",
      "author_url": "",
      "post_date": "2021-07-31T04:11:14.397000",
      "content": "<p>At first time, I use only 3-layers.<br>\nBecause model have more confused if we also used another 3-layers.</p>\n<p>However, In my personal opinion, there are some leakage in the off-channel.<br>\n(I am not sure about this) <br>\nMany values of the images looks like splited in on-channel and off-channel.</p>\n<p>like [0], [2], [4] channel corresponding with [3], [5], [1] channel. And It also can be flipped.<br>\nYou may easily found some images have more noise like Gaussian noise.<br>\nI think that is not guassian noise, just a scar of spliting image in 2 channel.</p>\n<p>There are total 48 combinations for recover original images.<br>\nEven though there are some leakage on off-channel, <br>\nwe can't directly using it without knowing exact information.<br>\nIt just my personal opinion.</p>\n<p>This competition like to find a steganography…</p>",
      "votes": 2,
      "replies": [
        {
          "id": 1406022,
          "author_name": "Robert Kim",
          "author_url": "",
          "post_date": "2021-07-31T12:17:00.063000",
          "content": "<p>Sometimes \"off\" signals look similar to the \"on\" signals because most probably man-made noise was present in both signals. </p>\n<p>Not sure if it could be considered as a leakage tho, on the contrary, model may look at the difference between on and off signal and infer whether signal present in the on and not present in the off can be considered \"alien\"</p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1445959,
      "author_name": "Araik Tamazian",
      "author_url": "",
      "post_date": "2021-08-04T10:36:20.793000",
      "content": "<p>I also use only 3 layers (\"ON\").</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1441320": "For me using \"ON\" observation i.e [0,2,4] works better than using all six channels",
    "1405578": "At first time, I use only 3-layers.\nBecause model have more confused if we also used another 3-layers.\n\nHowever, In my personal opinion, there are some leakage in the off-channel.\n(I am not sure about this) \nMany values of the images looks like splited in on-channel and off-channel.\n\nlike [0], [2], [4] channel corresponding with [3], [5], [1] channel. And It also can be flipped.\nYou may easily found some images have more noise like Gaussian noise.\nI think that is not guassian noise, just a scar of spliting image in 2 channel.\n\nThere are total 48 combinations for recover original images.\nEven though there are some leakage on off-channel, \nwe can't directly using it without knowing exact information.\nIt just my personal opinion.\n\nThis competition like to find a steganography...",
    "1405213": "I see that some people use all 6 layers. Someone uses 3 layers. My best results are obtained with 3 layers with target. How much do you use?",
    "1445959": "I also use only 3 layers (\"ON\")."
  }
}