{
  "id": 181247,
  "title": "Obtaining models from tensorflow website and a resource for the competition",
  "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/181247",
  "author_name": "",
  "post_date": "2020-09-08T05:48:53.374521600Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>All,</p>\n<p>I see that there are quite a few kernels which seems to utilize tensorflow models which more than likely come from the following website:</p>\n<p><a href=\"https://tfhub.dev/s?module-type=image-augmentation,image-classification,image-feature-vector,image-generator,image-object-detection,image-others,image-style-transfer,image-rnn-agent\" target=\"_blank\">https://tfhub.dev/s?module-type=image-augmentation,image-classification,image-feature-vector,image-generator,image-object-detection,image-others,image-style-transfer,image-rnn-agent</a></p>\n<p>You can also go to the above website by clicking on the following link and then click \"image\" in the section of \"browse by problem domain.\"</p>\n<p><a href=\"https://tfhub.dev/\" target=\"_blank\">https://tfhub.dev/</a></p>\n<p>However, my problems seems to stem from getting from the information found in these models (in the .gz file format) to then convert it to a jupyter file.  I have attached a jpeg picture to this post of what these files look like from the website above after I have downloaded these models onto my computer.</p>\n<p>Does anyone know how to extract the python code from these .gz files so that I can then utilize the models into the competition?  I have been stuck on that all day and I need some help.</p>\n<p>The above website can also be a resource for those who are not aware of it.</p>\n<p>Thank you for your future assistance.</p>\n<p>Brian</p>",
  "messages": [
    {
      "id": "1002425",
      "postDate": "09/08/2020 05:48:53",
      "content": "<p>All,</p>\n<p>I see that there are quite a few kernels which seems to utilize tensorflow models which more than likely come from the following website:</p>\n<p><a href=\"https://tfhub.dev/s?module-type=image-augmentation,image-classification,image-feature-vector,image-generator,image-object-detection,image-others,image-style-transfer,image-rnn-agent\" target=\"_blank\">https://tfhub.dev/s?module-type=image-augmentation,image-classification,image-feature-vector,image-generator,image-object-detection,image-others,image-style-transfer,image-rnn-agent</a></p>\n<p>You can also go to the above website by clicking on the following link and then click \"image\" in the section of \"browse by problem domain.\"</p>\n<p><a href=\"https://tfhub.dev/\" target=\"_blank\">https://tfhub.dev/</a></p>\n<p>However, my problems seems to stem from getting from the information found in these models (in the .gz file format) to then convert it to a jupyter file.  I have attached a jpeg picture to this post of what these files look like from the website above after I have downloaded these models onto my computer.</p>\n<p>Does anyone know how to extract the python code from these .gz files so that I can then utilize the models into the competition?  I have been stuck on that all day and I need some help.</p>\n<p>The above website can also be a resource for those who are not aware of it.</p>\n<p>Thank you for your future assistance.</p>\n<p>Brian</p>",
      "rawMarkdown": "All,\n\nI see that there are quite a few kernels which seems to utilize tensorflow models which more than likely come from the following website:\n\nhttps://tfhub.dev/s?module-type=image-augmentation,image-classification,image-feature-vector,image-generator,image-object-detection,image-others,image-style-transfer,image-rnn-agent\n\nYou can also go to the above website by clicking on the following link and then click \"image\" in the section of \"browse by problem domain.\"\n\nhttps://tfhub.dev/\n\nHowever, my problems seems to stem from getting from the information found in these models (in the .gz file format) to then convert it to a jupyter file.  I have attached a jpeg picture to this post of what these files look like from the website above after I have downloaded these models onto my computer.\n\nDoes anyone know how to extract the python code from these .gz files so that I can then utilize the models into the competition?  I have been stuck on that all day and I need some help.\n\nThe above website can also be a resource for those who are not aware of it.\n\nThank you for your future assistance.\n\nBrian",
      "votes": null
    },
    {
      "id": "1002569",
      "postDate": "09/08/2020 08:20:10",
      "content": "<p><a href=\"https://www.kaggle.com/beamers\" target=\"_blank\">@beamers</a> Those are model weights or feature vectors that contain coefficients or weights which have been compressed. Even if you unzip the compressed files there is no python code within those files for you to extract. To use those model files download and add it to kaggle as a dataset, then load the file using TF2 which is mentioned on their guide on the same page.</p>",
      "rawMarkdown": "beamers Those are model weights or feature vectors that contain coefficients or weights which have been compressed. Even if you unzip the compressed files there is no python code within those files for you to extract. To use those model files download and add it to kaggle as a dataset, then load the file using TF2 which is mentioned on their guide on the same page.",
      "votes": null
    },
    {
      "id": "1002959",
      "postDate": "09/08/2020 14:57:15",
      "content": "<p>Thank you for your response.  I am still learning and that has helped a lot.</p>",
      "rawMarkdown": "Thank you for your response.  I am still learning and that has helped a lot.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1002569,
      "author_name": "yovinyahathugoda",
      "author_url": "",
      "post_date": "09/08/2020 08:20:10",
      "content": "<p><a href=\"https://www.kaggle.com/beamers\" target=\"_blank\">@beamers</a> Those are model weights or feature vectors that contain coefficients or weights which have been compressed. Even if you unzip the compressed files there is no python code within those files for you to extract. To use those model files download and add it to kaggle as a dataset, then load the file using TF2 which is mentioned on their guide on the same page.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1002959,
          "author_name": "beamers",
          "author_url": "",
          "post_date": "09/08/2020 14:57:15",
          "content": "<p>Thank you for your response.  I am still learning and that has helped a lot.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1002425": "All,\n\nI see that there are quite a few kernels which seems to utilize tensorflow models which more than likely come from the following website:\n\nhttps://tfhub.dev/s?module-type=image-augmentation,image-classification,image-feature-vector,image-generator,image-object-detection,image-others,image-style-transfer,image-rnn-agent\n\nYou can also go to the above website by clicking on the following link and then click \"image\" in the section of \"browse by problem domain.\"\n\nhttps://tfhub.dev/\n\nHowever, my problems seems to stem from getting from the information found in these models (in the .gz file format) to then convert it to a jupyter file.  I have attached a jpeg picture to this post of what these files look like from the website above after I have downloaded these models onto my computer.\n\nDoes anyone know how to extract the python code from these .gz files so that I can then utilize the models into the competition?  I have been stuck on that all day and I need some help.\n\nThe above website can also be a resource for those who are not aware of it.\n\nThank you for your future assistance.\n\nBrian",
    "1002569": "beamers Those are model weights or feature vectors that contain coefficients or weights which have been compressed. Even if you unzip the compressed files there is no python code within those files for you to extract. To use those model files download and add it to kaggle as a dataset, then load the file using TF2 which is mentioned on their guide on the same page.",
    "1002959": "Thank you for your response.  I am still learning and that has helped a lot."
  },
  "source": "meta"
}