{
  "id": 415582,
  "title": "I think I actually see what ink is.",
  "url": "/competitions/vesuvius-challenge-ink-detection/discussion/415582",
  "author_name": "",
  "post_date": "2023-06-07T05:47:53.396325Z",
  "votes": 5,
  "comment_count": 1,
  "views": 0,
  "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2586962%2Fa3db5cf9005136c08e55182f1597321f%2FScreenshot%202023-06-06%20at%2010.27.33%20PM.png?generation=1686116290427292&amp;alt=media\" alt=\"\"></p>\n<p>Today I finally was able to visually correlate one pattern seen in CT data with areas labelled as ink and I'm actually thinking it might be how carbonized ink looks like on CT scans.</p>\n<p>I was looking at training sample 3, for which less visible ink on IR image suggest ink may lay a little deeper then on other fragments. So image above is from one of the layers 26-30, right where letters \"oyk`\" are on the label.</p>\n<p>do you see those overall lighter and smoother areas with sharp dark cracks over them? Those coinside with ink pretty well. So is it what might be called carbonized ink?</p>\n<p>Thoughts?</p>",
  "messages": [
    {
      "id": "2290802",
      "postDate": "06/07/2023 05:47:53",
      "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2586962%2Fa3db5cf9005136c08e55182f1597321f%2FScreenshot%202023-06-06%20at%2010.27.33%20PM.png?generation=1686116290427292&amp;alt=media\" alt=\"\"></p>\n<p>Today I finally was able to visually correlate one pattern seen in CT data with areas labelled as ink and I'm actually thinking it might be how carbonized ink looks like on CT scans.</p>\n<p>I was looking at training sample 3, for which less visible ink on IR image suggest ink may lay a little deeper then on other fragments. So image above is from one of the layers 26-30, right where letters \"oyk`\" are on the label.</p>\n<p>do you see those overall lighter and smoother areas with sharp dark cracks over them? Those coinside with ink pretty well. So is it what might be called carbonized ink?</p>\n<p>Thoughts?</p>",
      "rawMarkdown": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2586962%2Fa3db5cf9005136c08e55182f1597321f%2FScreenshot%202023-06-06%20at%2010.27.33%20PM.png?generation=1686116290427292&alt=media)\n\nToday I finally was able to visually correlate one pattern seen in CT data with areas labelled as ink and I'm actually thinking it might be how carbonized ink looks like on CT scans.\n\nI was looking at training sample 3, for which less visible ink on IR image suggest ink may lay a little deeper then on other fragments. So image above is from one of the layers 26-30, right where letters \"oyk`\" are on the label.\n\ndo you see those overall lighter and smoother areas with sharp dark cracks over them? Those coinside with ink pretty well. So is it what might be called carbonized ink?\n\nThoughts?",
      "votes": null
    },
    {
      "id": "2291879",
      "postDate": "06/07/2023 22:40:37",
      "content": "<p>Well, if this is the case, I suspect utilizing histogram layers can give better results then traditional CNN can do because we are dealing with statistical rather then geometrical features here. See e.g. <a href=\"https://www.joshpeeples.com/Histogram_Layer/\" target=\"_blank\">https://www.joshpeeples.com/Histogram_Layer/</a></p>",
      "rawMarkdown": "Well, if this is the case, I suspect utilizing histogram layers can give better results then traditional CNN can do because we are dealing with statistical rather then geometrical features here. See e.g. https://www.joshpeeples.com/Histogram_Layer/",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2291879,
      "author_name": "elvenmonk",
      "author_url": "",
      "post_date": "06/07/2023 22:40:37",
      "content": "<p>Well, if this is the case, I suspect utilizing histogram layers can give better results then traditional CNN can do because we are dealing with statistical rather then geometrical features here. See e.g. <a href=\"https://www.joshpeeples.com/Histogram_Layer/\" target=\"_blank\">https://www.joshpeeples.com/Histogram_Layer/</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2290802": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2586962%2Fa3db5cf9005136c08e55182f1597321f%2FScreenshot%202023-06-06%20at%2010.27.33%20PM.png?generation=1686116290427292&alt=media)\n\nToday I finally was able to visually correlate one pattern seen in CT data with areas labelled as ink and I'm actually thinking it might be how carbonized ink looks like on CT scans.\n\nI was looking at training sample 3, for which less visible ink on IR image suggest ink may lay a little deeper then on other fragments. So image above is from one of the layers 26-30, right where letters \"oyk`\" are on the label.\n\ndo you see those overall lighter and smoother areas with sharp dark cracks over them? Those coinside with ink pretty well. So is it what might be called carbonized ink?\n\nThoughts?",
    "2291879": "Well, if this is the case, I suspect utilizing histogram layers can give better results then traditional CNN can do because we are dealing with statistical rather then geometrical features here. See e.g. https://www.joshpeeples.com/Histogram_Layer/"
  },
  "source": "meta"
}