{
  "id": 612845,
  "title": "Test image segments",
  "url": "/competitions/physionet-ecg-image-digitization/discussion/612845",
  "author_name": "",
  "post_date": "2025-10-22T14:50:39.310186Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi,<br>\nAre the test images an Original color ECG image or could it be any of the segments (printed or mobile photo as well)? In short, do the models we train have to be robust across these segments? The original color image is cleaner and more consistent than the other segments. </p>",
  "messages": [
    {
      "id": "3305346",
      "postDate": "10/22/2025 14:50:39",
      "content": "<p>Hi,<br>\nAre the test images an Original color ECG image or could it be any of the segments (printed or mobile photo as well)? In short, do the models we train have to be robust across these segments? The original color image is cleaner and more consistent than the other segments. </p>",
      "rawMarkdown": "Hi,\nAre the test images an Original color ECG image or could it be any of the segments (printed or mobile photo as well)? In short, do the models we train have to be robust across these segments? The original color image is cleaner and more consistent than the other segments.",
      "votes": null
    },
    {
      "id": "3305351",
      "postDate": "10/22/2025 14:57:21",
      "content": "<p>The test images are mixtures of all the variants (printed, distorted and scanned, mobile photos, etc.)</p>",
      "rawMarkdown": "The test images are mixtures of all the variants (printed, distorted and scanned, mobile photos, etc.)",
      "votes": null
    },
    {
      "id": "3305355",
      "postDate": "10/22/2025 15:01:43",
      "content": "<p>Like the ECG images in the training data, the ECG images in the test data are different (scans of papers, photos of papers, photos of computer monitors).</p>\n<p>As you noted, some of these variants be easier to digitize than others, but we need to digitize the provided ECG images, not our preferred ECG images. </p>",
      "rawMarkdown": "Like the ECG images in the training data, the ECG images in the test data are different (scans of papers, photos of papers, photos of computer monitors).\n\nAs you noted, some of these variants be easier to digitize than others, but we need to digitize the provided ECG images, not our preferred ECG images.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3305351,
      "author_name": "r2241272",
      "author_url": "",
      "post_date": "10/22/2025 14:57:21",
      "content": "<p>The test images are mixtures of all the variants (printed, distorted and scanned, mobile photos, etc.)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3305355,
      "author_name": "matthewreyna",
      "author_url": "",
      "post_date": "10/22/2025 15:01:43",
      "content": "<p>Like the ECG images in the training data, the ECG images in the test data are different (scans of papers, photos of papers, photos of computer monitors).</p>\n<p>As you noted, some of these variants be easier to digitize than others, but we need to digitize the provided ECG images, not our preferred ECG images. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3305346": "Hi,\nAre the test images an Original color ECG image or could it be any of the segments (printed or mobile photo as well)? In short, do the models we train have to be robust across these segments? The original color image is cleaner and more consistent than the other segments.",
    "3305351": "The test images are mixtures of all the variants (printed, distorted and scanned, mobile photos, etc.)",
    "3305355": "Like the ECG images in the training data, the ECG images in the test data are different (scans of papers, photos of papers, photos of computer monitors).\n\nAs you noted, some of these variants be easier to digitize than others, but we need to digitize the provided ECG images, not our preferred ECG images."
  },
  "source": "meta"
}