{
  "id": 65794,
  "title": "this may be useful",
  "url": "/competitions/rsna-pneumonia-detection-challenge/discussion/65794",
  "author_name": "",
  "post_date": "2018-09-15T01:04:35.984126400Z",
  "votes": 9,
  "comment_count": 5,
  "views": 0,
  "content": "<p><a href=\"http://blog.qure.ai/notes/visualizing_deep_learning\">http://blog.qure.ai/notes/visualizing_deep_learning</a>\n  <img src=\"http://blog.qure.ai/assets/images/visualisation/xray.png\" alt=\"enter image description here\"></p>",
  "messages": [
    {
      "id": "387459",
      "postDate": "09/15/2018 01:04:35",
      "content": "<p><a href=\"http://blog.qure.ai/notes/visualizing_deep_learning\">http://blog.qure.ai/notes/visualizing_deep_learning</a>\n  <img src=\"http://blog.qure.ai/assets/images/visualisation/xray.png\" alt=\"enter image description here\"></p>",
      "rawMarkdown": "http://blog.qure.ai/notes/visualizing_deep_learning\n  ![enter image description here][1]\n\n\n  [1]: http://blog.qure.ai/assets/images/visualisation/xray.png",
      "votes": null
    },
    {
      "id": "387643",
      "postDate": "09/15/2018 10:41:35",
      "content": "<p>The might be helpful for understanding the theory of <a href=\"http://www.hackevolve.com/where-cnn-is-looking-grad-cam/\">grad-CAMs.</a><br>\ngithub code - <a href=\"https://github.com/jacobgil/pytorch-grad-cam/blob/master/grad-cam.py\">link</a><br>\nNew paper on Unet - <a href=\"https://github.com/SimonKohl/probabilistic_unet\">Probablistic Unet</a></p>",
      "rawMarkdown": "The might be helpful for understanding the theory of [grad-CAMs.][1]<br>\ngithub code - [link][2]<br>\nNew paper on Unet - [Probablistic Unet][3]\n\n  [1]: http://www.hackevolve.com/where-cnn-is-looking-grad-cam/\n[2]: https://github.com/jacobgil/pytorch-grad-cam/blob/master/grad-cam.py\n[3]: https://github.com/SimonKohl/probabilistic_unet",
      "votes": null
    },
    {
      "id": "387913",
      "postDate": "09/15/2018 22:40:40",
      "content": "<p>Note that this is not a feature of pneumonia but that of effusion.</p>",
      "rawMarkdown": "Note that this is not a feature of pneumonia but that of effusion.",
      "votes": null
    },
    {
      "id": "387995",
      "postDate": "09/16/2018 04:40:35",
      "content": "<p>Has anyone come across a case of pleural effusion with pneumonia? To be clear, I mean when the annotators labeled the pneumonia --not that there wasn't pneumonia present at all.</p>",
      "rawMarkdown": "Has anyone come across a case of pleural effusion with pneumonia? To be clear, I mean when the annotators labeled the pneumonia --not that there wasn't pneumonia present at all.",
      "votes": null
    },
    {
      "id": "390235",
      "postDate": "09/19/2018 22:31:09",
      "content": "<p>I would say they were annotating lung opacities and not necessarily pneumonia.</p>",
      "rawMarkdown": "I would say they were annotating lung opacities and not necessarily pneumonia.",
      "votes": null
    },
    {
      "id": "391139",
      "postDate": "09/21/2018 10:16:30",
      "content": "<p>In my understanding, the effusion is not regarded as opacity in the train data. It is because that in <a href=\"https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64723\">this topic</a> the organizer said that consolidation, infiltration and pneumonia label of  NIH Chest X-ray 14 dataset are regarded as opacity but effusion is not. One more evidence, in NIH Chest X-ray 14 dataset there are some bounding boxes of effusion. Here I show you some of them. Red bounding boxes are effusion bounding box from NIH Chest X-ray14 dataset and green bounding box are opacity bounding box from the train data.<img src=\"https://i.imgur.com/od4EsNb.png\" alt=\"effusion bounding box\"></p>",
      "rawMarkdown": "In my understanding, the effusion is not regarded as opacity in the train data. It is because that in [this topic](https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64723) the organizer said that consolidation, infiltration and pneumonia label of  NIH Chest X-ray 14 dataset are regarded as opacity but effusion is not. One more evidence, in NIH Chest X-ray 14 dataset there are some bounding boxes of effusion. Here I show you some of them. Red bounding boxes are effusion bounding box from NIH Chest X-ray14 dataset and green bounding box are opacity bounding box from the train data.![effusion bounding box][1]\n\n\n  [1]: https://i.imgur.com/od4EsNb.png",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 387643,
      "author_name": "trekkerthemaker",
      "author_url": "",
      "post_date": "09/15/2018 10:41:35",
      "content": "<p>The might be helpful for understanding the theory of <a href=\"http://www.hackevolve.com/where-cnn-is-looking-grad-cam/\">grad-CAMs.</a><br>\ngithub code - <a href=\"https://github.com/jacobgil/pytorch-grad-cam/blob/master/grad-cam.py\">link</a><br>\nNew paper on Unet - <a href=\"https://github.com/SimonKohl/probabilistic_unet\">Probablistic Unet</a></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 387913,
      "author_name": "osciiart",
      "author_url": "",
      "post_date": "09/15/2018 22:40:40",
      "content": "<p>Note that this is not a feature of pneumonia but that of effusion.</p>",
      "votes": null,
      "replies": [
        {
          "id": 387995,
          "author_name": "puremath86",
          "author_url": "",
          "post_date": "09/16/2018 04:40:35",
          "content": "<p>Has anyone come across a case of pleural effusion with pneumonia? To be clear, I mean when the annotators labeled the pneumonia --not that there wasn't pneumonia present at all.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 390235,
          "author_name": "arpandhatt",
          "author_url": "",
          "post_date": "09/19/2018 22:31:09",
          "content": "<p>I would say they were annotating lung opacities and not necessarily pneumonia.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 391139,
          "author_name": "osciiart",
          "author_url": "",
          "post_date": "09/21/2018 10:16:30",
          "content": "<p>In my understanding, the effusion is not regarded as opacity in the train data. It is because that in <a href=\"https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64723\">this topic</a> the organizer said that consolidation, infiltration and pneumonia label of  NIH Chest X-ray 14 dataset are regarded as opacity but effusion is not. One more evidence, in NIH Chest X-ray 14 dataset there are some bounding boxes of effusion. Here I show you some of them. Red bounding boxes are effusion bounding box from NIH Chest X-ray14 dataset and green bounding box are opacity bounding box from the train data.<img src=\"https://i.imgur.com/od4EsNb.png\" alt=\"effusion bounding box\"></p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "387459": "http://blog.qure.ai/notes/visualizing_deep_learning\n  ![enter image description here][1]\n\n\n  [1]: http://blog.qure.ai/assets/images/visualisation/xray.png",
    "387643": "The might be helpful for understanding the theory of [grad-CAMs.][1]<br>\ngithub code - [link][2]<br>\nNew paper on Unet - [Probablistic Unet][3]\n\n  [1]: http://www.hackevolve.com/where-cnn-is-looking-grad-cam/\n[2]: https://github.com/jacobgil/pytorch-grad-cam/blob/master/grad-cam.py\n[3]: https://github.com/SimonKohl/probabilistic_unet",
    "387913": "Note that this is not a feature of pneumonia but that of effusion.",
    "387995": "Has anyone come across a case of pleural effusion with pneumonia? To be clear, I mean when the annotators labeled the pneumonia --not that there wasn't pneumonia present at all.",
    "390235": "I would say they were annotating lung opacities and not necessarily pneumonia.",
    "391139": "In my understanding, the effusion is not regarded as opacity in the train data. It is because that in [this topic](https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64723) the organizer said that consolidation, infiltration and pneumonia label of  NIH Chest X-ray 14 dataset are regarded as opacity but effusion is not. One more evidence, in NIH Chest X-ray 14 dataset there are some bounding boxes of effusion. Here I show you some of them. Red bounding boxes are effusion bounding box from NIH Chest X-ray14 dataset and green bounding box are opacity bounding box from the train data.![effusion bounding box][1]\n\n\n  [1]: https://i.imgur.com/od4EsNb.png"
  },
  "source": "meta"
}