{
  "id": 75018,
  "title": "How to obtain se resnext-50, 100 weight for keras",
  "url": "/competitions/human-protein-atlas-image-classification/discussion/75018",
  "author_name": "",
  "post_date": "2018-12-18T01:03:02.146600600Z",
  "votes": 2,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I want to use se resnext-50, 100. And I found the github <a href=\"https://github.com/titu1994/keras-squeeze-excite-network/blob/master/se_resnext.py\">https://github.com/titu1994/keras-squeeze-excite-network/blob/master/se_resnext.py</a>.</p>\n\n<p>But, pretrained-weights is not included. </p>\n\n<p>How can I obtain the weights of se resnext-50, 100 for keras? please, help me!</p>",
  "messages": [
    {
      "id": "440768",
      "postDate": "12/18/2018 01:03:02",
      "content": "<p>I want to use se resnext-50, 100. And I found the github <a href=\"https://github.com/titu1994/keras-squeeze-excite-network/blob/master/se_resnext.py\">https://github.com/titu1994/keras-squeeze-excite-network/blob/master/se_resnext.py</a>.</p>\n\n<p>But, pretrained-weights is not included. </p>\n\n<p>How can I obtain the weights of se resnext-50, 100 for keras? please, help me!</p>",
      "rawMarkdown": "I want to use se resnext-50, 100. And I found the github https://github.com/titu1994/keras-squeeze-excite-network/blob/master/se_resnext.py.\n\nBut, pretrained-weights is not included. \n\nHow can I obtain the weights of se resnext-50, 100 for keras? please, help me!",
      "votes": null
    },
    {
      "id": "440797",
      "postDate": "12/18/2018 01:25:25",
      "content": "<p>There is something wrong with his implementation of resnext, you can see my ticket there.</p>",
      "rawMarkdown": "There is something wrong with his implementation of resnext, you can see my ticket there.",
      "votes": null
    },
    {
      "id": "442425",
      "postDate": "12/20/2018 00:16:24",
      "content": "<p>I tried titu's code but end up with very poor results. </p>",
      "rawMarkdown": "I tried titu's code but end up with very poor results.",
      "votes": null
    },
    {
      "id": "442476",
      "postDate": "12/20/2018 02:20:22",
      "content": "<p>Thanks for sharing. But, I want to know how to get 'se-resnext-50, 100', not resnet50. </p>\n\n<p>For now, I'm using the two model from scratch. Thanks!</p>",
      "rawMarkdown": "Thanks for sharing. But, I want to know how to get 'se-resnext-50, 100', not resnet50. \n\nFor now, I'm using the two model from scratch. Thanks!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 440797,
      "author_name": "moshel",
      "author_url": "",
      "post_date": "12/18/2018 01:25:25",
      "content": "<p>There is something wrong with his implementation of resnext, you can see my ticket there.</p>",
      "votes": null,
      "replies": [
        {
          "id": 442476,
          "author_name": "youhanlee",
          "author_url": "",
          "post_date": "12/20/2018 02:20:22",
          "content": "<p>Thanks for sharing. But, I want to know how to get 'se-resnext-50, 100', not resnet50. </p>\n\n<p>For now, I'm using the two model from scratch. Thanks!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 442425,
      "author_name": "kownse",
      "author_url": "",
      "post_date": "12/20/2018 00:16:24",
      "content": "<p>I tried titu's code but end up with very poor results. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "440768": "I want to use se resnext-50, 100. And I found the github https://github.com/titu1994/keras-squeeze-excite-network/blob/master/se_resnext.py.\n\nBut, pretrained-weights is not included. \n\nHow can I obtain the weights of se resnext-50, 100 for keras? please, help me!",
    "440797": "There is something wrong with his implementation of resnext, you can see my ticket there.",
    "442425": "I tried titu's code but end up with very poor results.",
    "442476": "Thanks for sharing. But, I want to know how to get 'se-resnext-50, 100', not resnet50. \n\nFor now, I'm using the two model from scratch. Thanks!"
  },
  "source": "meta"
}