{
  "id": 158260,
  "title": "Anyone tried to learn Cover vs Juniward model? ",
  "url": "/competitions/alaska2-image-steganalysis/discussion/158260",
  "author_name": "[ods.ai]Uladzimir Tumanau",
  "post_date": "2020-06-13T15:24:38.390000",
  "votes": 3,
  "comment_count": 7,
  "views": 0,
  "content": "<p>So far all my attempts has failed, because i get 0.5 accuracy which is random guessing. Any suggeations? What were your results?</p>",
  "messages": [
    {
      "id": 884726,
      "postDate": "2020-06-13T15:24:38.390Z",
      "content": "<p>So far all my attempts has failed, because i get 0.5 accuracy which is random guessing. Any suggeations? What were your results?</p>",
      "rawMarkdown": "So far all my attempts has failed, because i get 0.5 accuracy which is random guessing. Any suggeations? What were your results?",
      "votes": 3
    },
    {
      "id": 886534,
      "postDate": "2020-06-15T05:39:28.273Z",
      "content": "<p>Using pretrained model and small learning rate may help. I'm not sure.</p>",
      "rawMarkdown": "Using pretrained model and small learning rate may help. I'm not sure.",
      "votes": 1,
      "replies": [
        {
          "id": 889232,
          "postDate": "2020-06-16T20:59:28.170Z",
          "content": "<p>Have you tried to learn this yet? </p>",
          "rawMarkdown": "Have you tried to learn this yet? ",
          "votes": 1
        },
        {
          "id": 893848,
          "postDate": "2020-06-20T02:39:42.460Z",
          "content": "<p>Yes, It can get weighted AUC about 0.89.</p>",
          "rawMarkdown": "Yes, It can get weighted AUC about 0.89.",
          "votes": 2
        }
      ]
    },
    {
      "id": 886691,
      "postDate": "2020-06-15T07:49:25.510Z",
      "content": "<p>I was just thinking about this. I'll give it a try. Are you using a default pretrained net or using an already trained network with ALASKA dataset?</p>",
      "rawMarkdown": "I was just thinking about this. I'll give it a try. Are you using a default pretrained net or using an already trained network with ALASKA dataset?",
      "replies": [
        {
          "id": 889231,
          "postDate": "2020-06-16T20:59:10.287Z",
          "content": "<p>Already pretrained. Just finetuning</p>",
          "rawMarkdown": "Already pretrained. Just finetuning"
        },
        {
          "id": 889872,
          "postDate": "2020-06-17T07:36:30.920Z",
          "content": "<p>I also get bad results. Maybe using RGB only it's not the best approach to use for Juniward...</p>",
          "rawMarkdown": "I also get bad results. Maybe using RGB only it's not the best approach to use for Juniward..."
        }
      ]
    },
    {
      "id": 887073,
      "postDate": "2020-06-15T13:16:08.327Z",
      "rawMarkdown": "",
      "votes": 1,
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 886534,
      "author_name": "Johnny Lee",
      "author_url": "",
      "post_date": "2020-06-15T05:39:28.273000",
      "content": "<p>Using pretrained model and small learning rate may help. I'm not sure.</p>",
      "votes": 1,
      "replies": [
        {
          "id": 889232,
          "author_name": "[ods.ai]Uladzimir Tumanau",
          "author_url": "",
          "post_date": "2020-06-16T20:59:28.170000",
          "content": "<p>Have you tried to learn this yet? </p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 893848,
          "author_name": "Johnny Lee",
          "author_url": "",
          "post_date": "2020-06-20T02:39:42.460000",
          "content": "<p>Yes, It can get weighted AUC about 0.89.</p>",
          "votes": 2,
          "replies": []
        }
      ]
    },
    {
      "id": 886691,
      "author_name": "GuillemDelgado",
      "author_url": "",
      "post_date": "2020-06-15T07:49:25.510000",
      "content": "<p>I was just thinking about this. I'll give it a try. Are you using a default pretrained net or using an already trained network with ALASKA dataset?</p>",
      "votes": 0,
      "replies": [
        {
          "id": 889231,
          "author_name": "[ods.ai]Uladzimir Tumanau",
          "author_url": "",
          "post_date": "2020-06-16T20:59:10.287000",
          "content": "<p>Already pretrained. Just finetuning</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 889872,
          "author_name": "GuillemDelgado",
          "author_url": "",
          "post_date": "2020-06-17T07:36:30.920000",
          "content": "<p>I also get bad results. Maybe using RGB only it's not the best approach to use for Juniward...</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 887073,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-06-15T13:16:08.327000",
      "content": "",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "884726": "So far all my attempts has failed, because i get 0.5 accuracy which is random guessing. Any suggeations? What were your results?",
    "886534": "Using pretrained model and small learning rate may help. I'm not sure.",
    "886691": "I was just thinking about this. I'll give it a try. Are you using a default pretrained net or using an already trained network with ALASKA dataset?",
    "887073": ""
  }
}