{
  "id": 46757,
  "title": "Failure of the Classical approaches!!!",
  "url": "/competitions/sp-society-camera-model-identification/discussion/46757",
  "author_name": "",
  "post_date": "2018-01-02T20:58:16.644261100Z",
  "votes": null,
  "comment_count": 15,
  "views": 0,
  "content": "<p>Any classical approach we are trying seeming to fail in the test data. \nLike from 98% validation accuracy to 70% testing accuracy or lower!!! </p>\n\n<p>Is there anyone out there implementing classical approaches with any luck? </p>",
  "messages": [
    {
      "id": "264334",
      "postDate": "01/02/2018 20:58:16",
      "content": "<p>Any classical approach we are trying seeming to fail in the test data. \nLike from 98% validation accuracy to 70% testing accuracy or lower!!! </p>\n\n<p>Is there anyone out there implementing classical approaches with any luck? </p>",
      "rawMarkdown": "Any classical approach we are trying seeming to fail in the test data. \nLike from 98% validation accuracy to 70% testing accuracy or lower!!! \n\nIs there anyone out there implementing classical approaches with any luck?",
      "votes": null
    },
    {
      "id": "264847",
      "postDate": "01/04/2018 01:57:24",
      "content": "<p>So what was the Eureka moment to get you into 4th position ;)</p>",
      "rawMarkdown": "So what was the Eureka moment to get you into 4th position ;)",
      "votes": null
    },
    {
      "id": "264861",
      "postDate": "01/04/2018 02:45:14",
      "content": "<p>not doing things 'dropout at softmax guy' did ;) </p>",
      "rawMarkdown": "not doing things 'dropout at softmax guy' did ;)",
      "votes": null
    },
    {
      "id": "264987",
      "postDate": "01/04/2018 10:13:49",
      "content": "<p>Nice ;)</p>",
      "rawMarkdown": "Nice ;)",
      "votes": null
    },
    {
      "id": "265020",
      "postDate": "01/04/2018 12:10:09",
      "content": "<p>What did he do that you are not doing ? ( apart from applying dropout at the softmax ) </p>",
      "rawMarkdown": "What did he do that you are not doing ? ( apart from applying dropout at the softmax )",
      "votes": null
    },
    {
      "id": "265183",
      "postDate": "01/04/2018 20:12:52",
      "content": "<p>Out of interest CVxTz, is your best model or models derived from pre-trained networks like InceptionV3, Resnet50 etc or have you used your own CNN architecture? </p>",
      "rawMarkdown": "Out of interest CVxTz, is your best model or models derived from pre-trained networks like InceptionV3, Resnet50 etc or have you used your own CNN architecture?",
      "votes": null
    },
    {
      "id": "265190",
      "postDate": "01/04/2018 20:29:09",
      "content": "<p>Inception_resnet_v2 works the best for those I tried so far.</p>",
      "rawMarkdown": "Inception_resnet_v2 works the best for those I tried so far.",
      "votes": null
    },
    {
      "id": "265191",
      "postDate": "01/04/2018 20:30:24",
      "content": "<p>I find the same. Haven't submitted that yet though. I'm trying to solve the problem of test Vs validation discrepancy</p>",
      "rawMarkdown": "I find the same. Haven't submitted that yet though. I'm trying to solve the problem of test Vs validation discrepancy",
      "votes": null
    },
    {
      "id": "265193",
      "postDate": "01/04/2018 20:34:07",
      "content": "<p>Yeah, trying to find the trick too. Otherwise everything is useless :D </p>",
      "rawMarkdown": "Yeah, trying to find the trick too. Otherwise everything is useless :D",
      "votes": null
    },
    {
      "id": "265194",
      "postDate": "01/04/2018 20:35:00",
      "content": "<p>I have a few ideas if you'd like to try and team up? </p>",
      "rawMarkdown": "I have a few ideas if you'd like to try and team up?",
      "votes": null
    },
    {
      "id": "265195",
      "postDate": "01/04/2018 20:36:16",
      "content": "<p>Sure</p>",
      "rawMarkdown": "Sure",
      "votes": null
    },
    {
      "id": "265238",
      "postDate": "01/05/2018 00:05:30",
      "content": "<p>A competition master and a competition expert teaming up! :| Maybe I shouldn't have started this thread! </p>\n\n<p>Man this is getting really intense! :D </p>",
      "rawMarkdown": "A competition master and a competition expert teaming up! :| Maybe I shouldn't have started this thread! \n\nMan this is getting really intense! :D",
      "votes": null
    },
    {
      "id": "265239",
      "postDate": "01/05/2018 00:06:31",
      "content": "<p>Haha :) - A competition master and expert that aren't 'Subject matter experts' aren't anything to worry about ;)</p>",
      "rawMarkdown": "Haha :) - A competition master and expert that aren't 'Subject matter experts' aren't anything to worry about ;)",
      "votes": null
    },
    {
      "id": "265243",
      "postDate": "01/05/2018 00:24:45",
      "content": "<p>I hope what you are saying is true! :p </p>",
      "rawMarkdown": "I hope what you are saying is true! :p",
      "votes": null
    },
    {
      "id": "266913",
      "postDate": "01/10/2018 05:52:34",
      "content": "<p>What does the SPcup_eligible mean?</p>",
      "rawMarkdown": "What does the SPcup_eligible mean?",
      "votes": null
    },
    {
      "id": "267003",
      "postDate": "01/10/2018 10:48:14",
      "content": "<p>We are competing for the signal processing cup organized by IEEE simultaneously.\nTop 3 SP cup eligible teams will be going to Canada to solve a extended problem and the best among will win the SP cup 2018.\nOnly undergrad students are eligible for the SP cup. At most one Masters students is allowed in the team and they are advised by a professor of their university. </p>",
      "rawMarkdown": "We are competing for the signal processing cup organized by IEEE simultaneously.\nTop 3 SP cup eligible teams will be going to Canada to solve a extended problem and the best among will win the SP cup 2018.\nOnly undergrad students are eligible for the SP cup. At most one Masters students is allowed in the team and they are advised by a professor of their university.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 264847,
      "author_name": "craigglastonbury",
      "author_url": "",
      "post_date": "01/04/2018 01:57:24",
      "content": "<p>So what was the Eureka moment to get you into 4th position ;)</p>",
      "votes": null,
      "replies": [
        {
          "id": 264861,
          "author_name": "muntakimrafi",
          "author_url": "",
          "post_date": "01/04/2018 02:45:14",
          "content": "<p>not doing things 'dropout at softmax guy' did ;) </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 264987,
          "author_name": "craigglastonbury",
          "author_url": "",
          "post_date": "01/04/2018 10:13:49",
          "content": "<p>Nice ;)</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 265020,
          "author_name": "CVxTz",
          "author_url": "",
          "post_date": "01/04/2018 12:10:09",
          "content": "<p>What did he do that you are not doing ? ( apart from applying dropout at the softmax ) </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 265183,
          "author_name": "craigglastonbury",
          "author_url": "",
          "post_date": "01/04/2018 20:12:52",
          "content": "<p>Out of interest CVxTz, is your best model or models derived from pre-trained networks like InceptionV3, Resnet50 etc or have you used your own CNN architecture? </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 265190,
          "author_name": "CVxTz",
          "author_url": "",
          "post_date": "01/04/2018 20:29:09",
          "content": "<p>Inception_resnet_v2 works the best for those I tried so far.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 265191,
          "author_name": "craigglastonbury",
          "author_url": "",
          "post_date": "01/04/2018 20:30:24",
          "content": "<p>I find the same. Haven't submitted that yet though. I'm trying to solve the problem of test Vs validation discrepancy</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 265193,
          "author_name": "CVxTz",
          "author_url": "",
          "post_date": "01/04/2018 20:34:07",
          "content": "<p>Yeah, trying to find the trick too. Otherwise everything is useless :D </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 265194,
          "author_name": "craigglastonbury",
          "author_url": "",
          "post_date": "01/04/2018 20:35:00",
          "content": "<p>I have a few ideas if you'd like to try and team up? </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 265195,
          "author_name": "CVxTz",
          "author_url": "",
          "post_date": "01/04/2018 20:36:16",
          "content": "<p>Sure</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 265238,
          "author_name": "muntakimrafi",
          "author_url": "",
          "post_date": "01/05/2018 00:05:30",
          "content": "<p>A competition master and a competition expert teaming up! :| Maybe I shouldn't have started this thread! </p>\n\n<p>Man this is getting really intense! :D </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 265239,
          "author_name": "craigglastonbury",
          "author_url": "",
          "post_date": "01/05/2018 00:06:31",
          "content": "<p>Haha :) - A competition master and expert that aren't 'Subject matter experts' aren't anything to worry about ;)</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 265243,
          "author_name": "muntakimrafi",
          "author_url": "",
          "post_date": "01/05/2018 00:24:45",
          "content": "<p>I hope what you are saying is true! :p </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 266913,
      "author_name": "youngkl",
      "author_url": "",
      "post_date": "01/10/2018 05:52:34",
      "content": "<p>What does the SPcup_eligible mean?</p>",
      "votes": null,
      "replies": [
        {
          "id": 267003,
          "author_name": "muntakimrafi",
          "author_url": "",
          "post_date": "01/10/2018 10:48:14",
          "content": "<p>We are competing for the signal processing cup organized by IEEE simultaneously.\nTop 3 SP cup eligible teams will be going to Canada to solve a extended problem and the best among will win the SP cup 2018.\nOnly undergrad students are eligible for the SP cup. At most one Masters students is allowed in the team and they are advised by a professor of their university. </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "264334": "Any classical approach we are trying seeming to fail in the test data. \nLike from 98% validation accuracy to 70% testing accuracy or lower!!! \n\nIs there anyone out there implementing classical approaches with any luck?",
    "264847": "So what was the Eureka moment to get you into 4th position ;)",
    "264861": "not doing things 'dropout at softmax guy' did ;)",
    "264987": "Nice ;)",
    "265020": "What did he do that you are not doing ? ( apart from applying dropout at the softmax )",
    "265183": "Out of interest CVxTz, is your best model or models derived from pre-trained networks like InceptionV3, Resnet50 etc or have you used your own CNN architecture?",
    "265190": "Inception_resnet_v2 works the best for those I tried so far.",
    "265191": "I find the same. Haven't submitted that yet though. I'm trying to solve the problem of test Vs validation discrepancy",
    "265193": "Yeah, trying to find the trick too. Otherwise everything is useless :D",
    "265194": "I have a few ideas if you'd like to try and team up?",
    "265195": "Sure",
    "265238": "A competition master and a competition expert teaming up! :| Maybe I shouldn't have started this thread! \n\nMan this is getting really intense! :D",
    "265239": "Haha :) - A competition master and expert that aren't 'Subject matter experts' aren't anything to worry about ;)",
    "265243": "I hope what you are saying is true! :p",
    "266913": "What does the SPcup_eligible mean?",
    "267003": "We are competing for the signal processing cup organized by IEEE simultaneously.\nTop 3 SP cup eligible teams will be going to Canada to solve a extended problem and the best among will win the SP cup 2018.\nOnly undergrad students are eligible for the SP cup. At most one Masters students is allowed in the team and they are advised by a professor of their university."
  },
  "source": "meta"
}