{
  "id": 74068,
  "title": "One possible use for similar images",
  "url": "/competitions/human-protein-atlas-image-classification/discussion/74068",
  "author_name": "",
  "post_date": "2018-12-08T05:45:50.183364700Z",
  "votes": 3,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Not sure if anyone figured out a meaningful way of using similar images. When it comes to pairs of similar images found in test data, I suggest to use the better image for classification, and transfer the labels to the imperfect Siamese twin. FWIW, it appears that most of these differences are due to z-stacking.</p>\n\n<p>This is an example where both images are very good, but there is a pattern of green speckles in the cytosol that is more obvious on the right side. Nuclear speckles are stronger on the right as well.</p>\n\n<p><img src=\"https://i.postimg.cc/bwzyNj9T/test-similar-102.png\" alt=\"enter image description here\"></p>\n\n<p>This signal appears to be cytoplasmic, and maybe some in nucleus. I could never get the same information from the left image.</p>\n\n<p><img src=\"https://i.postimg.cc/s2MRvfQp/test-similar-108.png\" alt=\"enter image description here\"></p>\n\n<p>The right image is basically giberish.</p>\n\n<p><img src=\"https://i.postimg.cc/26vR7Hxy/test-similar-109.png\" alt=\"enter image description here\"></p>\n\n<p>The left image is just slightly better than giberish.</p>\n\n<p><img src=\"https://i.postimg.cc/YqtKQzL2/test-similar-111.png\" alt=\"enter image description here\"></p>\n\n<p>Not sure what is going on here, but it appears that microtubules on the right are obscuring nuclear signal.</p>\n\n<p><img src=\"https://i.postimg.cc/SxtBF0Tj/test-similar-112.png\" alt=\"enter image description here\"></p>\n\n<p>Wrong depth here, so microtubules on the right are obscuring a strong cytoplasmic signal.</p>\n\n<p><img src=\"https://i.postimg.cc/Vv1TBz6Z/test-similar-122.png\" alt=\"enter image description here\"></p>\n\n<p>Wrong depth again, this time obscuring a strong nuclear signal.</p>\n\n<p><img src=\"https://i.postimg.cc/90pkLswm/test-similar-123.png\" alt=\"enter image description here\"></p>\n\n<p>Man, the more I look at this last image on the right, the more it conjures the image of <a href=\"https://en.wikipedia.org/wiki/The_Phantom\"><strong>The Ghost Who Walks</strong></a>.</p>\n\n<p><img src=\"https://i.postimg.cc/4xn01LfG/81-L7-Ss-C4g-DL-SY500.jpg\" alt=\"enter image description here\"></p>\n\n<p>There is <a href=\"https://www.youtube.com/watch?v=8kUGJf04uHk\"><strong>a pretty good song</strong></a> by the same name.</p>",
  "messages": [
    {
      "id": "435496",
      "postDate": "12/08/2018 05:45:50",
      "content": "<p>Not sure if anyone figured out a meaningful way of using similar images. When it comes to pairs of similar images found in test data, I suggest to use the better image for classification, and transfer the labels to the imperfect Siamese twin. FWIW, it appears that most of these differences are due to z-stacking.</p>\n\n<p>This is an example where both images are very good, but there is a pattern of green speckles in the cytosol that is more obvious on the right side. Nuclear speckles are stronger on the right as well.</p>\n\n<p><img src=\"https://i.postimg.cc/bwzyNj9T/test-similar-102.png\" alt=\"enter image description here\"></p>\n\n<p>This signal appears to be cytoplasmic, and maybe some in nucleus. I could never get the same information from the left image.</p>\n\n<p><img src=\"https://i.postimg.cc/s2MRvfQp/test-similar-108.png\" alt=\"enter image description here\"></p>\n\n<p>The right image is basically giberish.</p>\n\n<p><img src=\"https://i.postimg.cc/26vR7Hxy/test-similar-109.png\" alt=\"enter image description here\"></p>\n\n<p>The left image is just slightly better than giberish.</p>\n\n<p><img src=\"https://i.postimg.cc/YqtKQzL2/test-similar-111.png\" alt=\"enter image description here\"></p>\n\n<p>Not sure what is going on here, but it appears that microtubules on the right are obscuring nuclear signal.</p>\n\n<p><img src=\"https://i.postimg.cc/SxtBF0Tj/test-similar-112.png\" alt=\"enter image description here\"></p>\n\n<p>Wrong depth here, so microtubules on the right are obscuring a strong cytoplasmic signal.</p>\n\n<p><img src=\"https://i.postimg.cc/Vv1TBz6Z/test-similar-122.png\" alt=\"enter image description here\"></p>\n\n<p>Wrong depth again, this time obscuring a strong nuclear signal.</p>\n\n<p><img src=\"https://i.postimg.cc/90pkLswm/test-similar-123.png\" alt=\"enter image description here\"></p>\n\n<p>Man, the more I look at this last image on the right, the more it conjures the image of <a href=\"https://en.wikipedia.org/wiki/The_Phantom\"><strong>The Ghost Who Walks</strong></a>.</p>\n\n<p><img src=\"https://i.postimg.cc/4xn01LfG/81-L7-Ss-C4g-DL-SY500.jpg\" alt=\"enter image description here\"></p>\n\n<p>There is <a href=\"https://www.youtube.com/watch?v=8kUGJf04uHk\"><strong>a pretty good song</strong></a> by the same name.</p>",
      "rawMarkdown": "Not sure if anyone figured out a meaningful way of using similar images. When it comes to pairs of similar images found in test data, I suggest to use the better image for classification, and transfer the labels to the imperfect Siamese twin. FWIW, it appears that most of these differences are due to z-stacking.\n\nThis is an example where both images are very good, but there is a pattern of green speckles in the cytosol that is more obvious on the right side. Nuclear speckles are stronger on the right as well.\n\n![enter image description here][1]\n\nThis signal appears to be cytoplasmic, and maybe some in nucleus. I could never get the same information from the left image.\n\n![enter image description here][2]\n\nThe right image is basically giberish.\n\n![enter image description here][3]\n\nThe left image is just slightly better than giberish.\n\n![enter image description here][4]\n\nNot sure what is going on here, but it appears that microtubules on the right are obscuring nuclear signal.\n\n![enter image description here][5]\n\nWrong depth here, so microtubules on the right are obscuring a strong cytoplasmic signal.\n\n![enter image description here][6]\n\nWrong depth again, this time obscuring a strong nuclear signal.\n\n![enter image description here][7]\n\nMan, the more I look at this last image on the right, the more it conjures the image of [__The Ghost Who Walks__](https://en.wikipedia.org/wiki/The_Phantom).\n\n![enter image description here][8]\n\nThere is [__a pretty good song__](https://www.youtube.com/watch?v=8kUGJf04uHk) by the same name.\n\n  [1]: https://i.postimg.cc/bwzyNj9T/test-similar-102.png\n  [2]: https://i.postimg.cc/s2MRvfQp/test-similar-108.png\n  [3]: https://i.postimg.cc/26vR7Hxy/test-similar-109.png\n  [4]: https://i.postimg.cc/YqtKQzL2/test-similar-111.png\n  [5]: https://i.postimg.cc/SxtBF0Tj/test-similar-112.png\n  [6]: https://i.postimg.cc/Vv1TBz6Z/test-similar-122.png\n  [7]: https://i.postimg.cc/90pkLswm/test-similar-123.png\n  [8]: https://i.postimg.cc/4xn01LfG/81-L7-Ss-C4g-DL-SY500.jpg",
      "votes": null
    },
    {
      "id": "435777",
      "postDate": "12/08/2018 18:51:02",
      "content": "<p>Congrats on becoming Discussion Grandmaster!  I knew it was a matter of time given how much you share.</p>",
      "rawMarkdown": "Congrats on becoming Discussion Grandmaster!  I knew it was a matter of time given how much you share.",
      "votes": null
    },
    {
      "id": "435784",
      "postDate": "12/08/2018 19:00:24",
      "content": "<p>Congrats! You definitely deserve it! </p>",
      "rawMarkdown": "Congrats! You definitely deserve it!",
      "votes": null
    },
    {
      "id": "435788",
      "postDate": "12/08/2018 19:19:42",
      "content": "<p>Thanks <a href=\"/cpmpml\">@cpmpml</a>, you and <a href=\"/hengck23\">@hengck23</a> set an example for all of us when it comes to sharing the knowledge.</p>",
      "rawMarkdown": "Thanks @cpmpml, you and @hengck23 set an example for all of us when it comes to sharing the knowledge.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 435777,
      "author_name": "cpmpml",
      "author_url": "",
      "post_date": "12/08/2018 18:51:02",
      "content": "<p>Congrats on becoming Discussion Grandmaster!  I knew it was a matter of time given how much you share.</p>",
      "votes": null,
      "replies": [
        {
          "id": 435788,
          "author_name": "tilii7",
          "author_url": "",
          "post_date": "12/08/2018 19:19:42",
          "content": "<p>Thanks <a href=\"/cpmpml\">@cpmpml</a>, you and <a href=\"/hengck23\">@hengck23</a> set an example for all of us when it comes to sharing the knowledge.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 435784,
      "author_name": "moshel",
      "author_url": "",
      "post_date": "12/08/2018 19:00:24",
      "content": "<p>Congrats! You definitely deserve it! </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "435496": "Not sure if anyone figured out a meaningful way of using similar images. When it comes to pairs of similar images found in test data, I suggest to use the better image for classification, and transfer the labels to the imperfect Siamese twin. FWIW, it appears that most of these differences are due to z-stacking.\n\nThis is an example where both images are very good, but there is a pattern of green speckles in the cytosol that is more obvious on the right side. Nuclear speckles are stronger on the right as well.\n\n![enter image description here][1]\n\nThis signal appears to be cytoplasmic, and maybe some in nucleus. I could never get the same information from the left image.\n\n![enter image description here][2]\n\nThe right image is basically giberish.\n\n![enter image description here][3]\n\nThe left image is just slightly better than giberish.\n\n![enter image description here][4]\n\nNot sure what is going on here, but it appears that microtubules on the right are obscuring nuclear signal.\n\n![enter image description here][5]\n\nWrong depth here, so microtubules on the right are obscuring a strong cytoplasmic signal.\n\n![enter image description here][6]\n\nWrong depth again, this time obscuring a strong nuclear signal.\n\n![enter image description here][7]\n\nMan, the more I look at this last image on the right, the more it conjures the image of [__The Ghost Who Walks__](https://en.wikipedia.org/wiki/The_Phantom).\n\n![enter image description here][8]\n\nThere is [__a pretty good song__](https://www.youtube.com/watch?v=8kUGJf04uHk) by the same name.\n\n  [1]: https://i.postimg.cc/bwzyNj9T/test-similar-102.png\n  [2]: https://i.postimg.cc/s2MRvfQp/test-similar-108.png\n  [3]: https://i.postimg.cc/26vR7Hxy/test-similar-109.png\n  [4]: https://i.postimg.cc/YqtKQzL2/test-similar-111.png\n  [5]: https://i.postimg.cc/SxtBF0Tj/test-similar-112.png\n  [6]: https://i.postimg.cc/Vv1TBz6Z/test-similar-122.png\n  [7]: https://i.postimg.cc/90pkLswm/test-similar-123.png\n  [8]: https://i.postimg.cc/4xn01LfG/81-L7-Ss-C4g-DL-SY500.jpg",
    "435777": "Congrats on becoming Discussion Grandmaster!  I knew it was a matter of time given how much you share.",
    "435784": "Congrats! You definitely deserve it!",
    "435788": "Thanks @cpmpml, you and @hengck23 set an example for all of us when it comes to sharing the knowledge."
  },
  "source": "meta"
}