{
  "id": 87749,
  "title": "Last years baseline",
  "url": "/competitions/inaturalist-2019-fgvc6/discussion/87749",
  "author_name": "",
  "post_date": "2019-04-03T05:05:30.377374700Z",
  "votes": 10,
  "comment_count": 9,
  "views": 0,
  "content": "<p>Hey everyone,</p>\n\n<p>I have been considering to join this competition and did some research on the dataset and past solutions. I stumbled upon the baseline model from last years iNatural competition and thought I'd share it here:</p>\n\n<p><a href=\"https://github.com/macaodha/inat_comp_2018\">https://github.com/macaodha/inat_comp_2018</a></p>\n\n<p>As the dataset for this year is much different from the one this model was created for, it might not give any good results, but maybe somebody still finds this helpful!</p>",
  "messages": [
    {
      "id": "506199",
      "postDate": "04/03/2019 05:05:30",
      "content": "<p>Hey everyone,</p>\n\n<p>I have been considering to join this competition and did some research on the dataset and past solutions. I stumbled upon the baseline model from last years iNatural competition and thought I'd share it here:</p>\n\n<p><a href=\"https://github.com/macaodha/inat_comp_2018\">https://github.com/macaodha/inat_comp_2018</a></p>\n\n<p>As the dataset for this year is much different from the one this model was created for, it might not give any good results, but maybe somebody still finds this helpful!</p>",
      "rawMarkdown": "Hey everyone,\n\nI have been considering to join this competition and did some research on the dataset and past solutions. I stumbled upon the baseline model from last years iNatural competition and thought I'd share it here:\n\n[https://github.com/macaodha/inat_comp_2018](https://github.com/macaodha/inat_comp_2018)\n\nAs the dataset for this year is much different from the one this model was created for, it might not give any good results, but maybe somebody still finds this helpful!",
      "votes": null
    },
    {
      "id": "506253",
      "postDate": "04/03/2019 07:46:37",
      "content": "<p>Hi I just posted a kernel with some attempts to get this working, advice welcome. Thanks!</p>",
      "rawMarkdown": "Hi I just posted a kernel with some attempts to get this working, advice welcome. Thanks!",
      "votes": null
    },
    {
      "id": "506270",
      "postDate": "04/03/2019 08:22:20",
      "content": "<p>A slight change( I have newer version  of pytorch, so I changed a little bit syntax) on this will give at least 0.22 on public lead-board, I have difficulty on pushing forward, going to other direction now. Maybe old data is useful but unfortunately I need to re-download them. </p>",
      "rawMarkdown": "A slight change( I have newer version  of pytorch, so I changed a little bit syntax) on this will give at least 0.22 on public lead-board, I have difficulty on pushing forward, going to other direction now. Maybe old data is useful but unfortunately I need to re-download them.",
      "votes": null
    },
    {
      "id": "506640",
      "postDate": "04/03/2019 17:56:44",
      "content": "<p>Nice work zz, did you use the 2018 inception weights?</p>",
      "rawMarkdown": "Nice work zz, did you use the 2018 inception weights?",
      "votes": null
    },
    {
      "id": "506719",
      "postDate": "04/03/2019 19:44:43",
      "content": "<p>no, using image-net pretrained model</p>",
      "rawMarkdown": "no, using image-net pretrained model",
      "votes": null
    },
    {
      "id": "531455",
      "postDate": "05/15/2019 00:11:55",
      "content": "<p>Hi <a href=\"/weizhezhang\">@weizhezhang</a> may I ask how many epochs you reach to 0.22?</p>",
      "rawMarkdown": "Hi @weizhezhang may I ask how many epochs you reach to 0.22?",
      "votes": null
    },
    {
      "id": "531748",
      "postDate": "05/15/2019 13:24:46",
      "content": "<p>I was out of this competition, it seems like 20 or 30 </p>",
      "rawMarkdown": "I was out of this competition, it seems like 20 or 30",
      "votes": null
    },
    {
      "id": "531959",
      "postDate": "05/15/2019 23:30:15",
      "content": "<p>Thank you <a href=\"/weizhezhang\">@weizhezhang</a> :)! To reach epochs 20 or 30 will be a problem train on Kaggle</p>",
      "rawMarkdown": "Thank you @weizhezhang :)! To reach epochs 20 or 30 will be a problem train on Kaggle",
      "votes": null
    },
    {
      "id": "532634",
      "postDate": "05/17/2019 12:33:37",
      "content": "<p>save the best checkpoint and continue next launch</p>",
      "rawMarkdown": "save the best checkpoint and continue next launch",
      "votes": null
    },
    {
      "id": "532863",
      "postDate": "05/17/2019 23:13:49",
      "content": "<p><a href=\"/weizhezhang\">@weizhezhang</a> Oh right! Appreciate your great advice:)!</p>",
      "rawMarkdown": "weizhezhang Oh right! Appreciate your great advice:)!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 506253,
      "author_name": "interneuron",
      "author_url": "",
      "post_date": "04/03/2019 07:46:37",
      "content": "<p>Hi I just posted a kernel with some attempts to get this working, advice welcome. Thanks!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 506270,
      "author_name": "weizhezhang",
      "author_url": "",
      "post_date": "04/03/2019 08:22:20",
      "content": "<p>A slight change( I have newer version  of pytorch, so I changed a little bit syntax) on this will give at least 0.22 on public lead-board, I have difficulty on pushing forward, going to other direction now. Maybe old data is useful but unfortunately I need to re-download them. </p>",
      "votes": null,
      "replies": [
        {
          "id": 506640,
          "author_name": "interneuron",
          "author_url": "",
          "post_date": "04/03/2019 17:56:44",
          "content": "<p>Nice work zz, did you use the 2018 inception weights?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 506719,
          "author_name": "weizhezhang",
          "author_url": "",
          "post_date": "04/03/2019 19:44:43",
          "content": "<p>no, using image-net pretrained model</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 531455,
          "author_name": "hsinwenchang",
          "author_url": "",
          "post_date": "05/15/2019 00:11:55",
          "content": "<p>Hi <a href=\"/weizhezhang\">@weizhezhang</a> may I ask how many epochs you reach to 0.22?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 531748,
          "author_name": "weizhezhang",
          "author_url": "",
          "post_date": "05/15/2019 13:24:46",
          "content": "<p>I was out of this competition, it seems like 20 or 30 </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 531959,
          "author_name": "hsinwenchang",
          "author_url": "",
          "post_date": "05/15/2019 23:30:15",
          "content": "<p>Thank you <a href=\"/weizhezhang\">@weizhezhang</a> :)! To reach epochs 20 or 30 will be a problem train on Kaggle</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 532634,
          "author_name": "weizhezhang",
          "author_url": "",
          "post_date": "05/17/2019 12:33:37",
          "content": "<p>save the best checkpoint and continue next launch</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 532863,
          "author_name": "hsinwenchang",
          "author_url": "",
          "post_date": "05/17/2019 23:13:49",
          "content": "<p><a href=\"/weizhezhang\">@weizhezhang</a> Oh right! Appreciate your great advice:)!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "506199": "Hey everyone,\n\nI have been considering to join this competition and did some research on the dataset and past solutions. I stumbled upon the baseline model from last years iNatural competition and thought I'd share it here:\n\n[https://github.com/macaodha/inat_comp_2018](https://github.com/macaodha/inat_comp_2018)\n\nAs the dataset for this year is much different from the one this model was created for, it might not give any good results, but maybe somebody still finds this helpful!",
    "506253": "Hi I just posted a kernel with some attempts to get this working, advice welcome. Thanks!",
    "506270": "A slight change( I have newer version  of pytorch, so I changed a little bit syntax) on this will give at least 0.22 on public lead-board, I have difficulty on pushing forward, going to other direction now. Maybe old data is useful but unfortunately I need to re-download them.",
    "506640": "Nice work zz, did you use the 2018 inception weights?",
    "506719": "no, using image-net pretrained model",
    "531455": "Hi @weizhezhang may I ask how many epochs you reach to 0.22?",
    "531748": "I was out of this competition, it seems like 20 or 30",
    "531959": "Thank you @weizhezhang :)! To reach epochs 20 or 30 will be a problem train on Kaggle",
    "532634": "save the best checkpoint and continue next launch",
    "532863": "weizhezhang Oh right! Appreciate your great advice:)!"
  },
  "source": "meta"
}