{
  "id": 402415,
  "title": "should I stack all 65 images or a selected portion in between?",
  "url": "/competitions/vesuvius-challenge-ink-detection/discussion/402415",
  "author_name": "",
  "post_date": "2023-04-18T09:51:33.961501300Z",
  "votes": 4,
  "comment_count": 5,
  "views": 0,
  "content": "<p>Since I have limited resources, which include only Kaggle and Colab, concatenating 65 images has become a challenge due to the exceeding RAM memory. To overcome this, I am currently reducing the size of each image by a third and converting them into uint8 format. or should I take a segment of slices from in between </p>",
  "messages": [
    {
      "id": "2225608",
      "postDate": "04/18/2023 09:51:33",
      "content": "<p>Since I have limited resources, which include only Kaggle and Colab, concatenating 65 images has become a challenge due to the exceeding RAM memory. To overcome this, I am currently reducing the size of each image by a third and converting them into uint8 format. or should I take a segment of slices from in between </p>",
      "rawMarkdown": "Since I have limited resources, which include only Kaggle and Colab, concatenating 65 images has become a challenge due to the exceeding RAM memory. To overcome this, I am currently reducing the size of each image by a third and converting them into uint8 format. or should I take a segment of slices from in between",
      "votes": null
    },
    {
      "id": "2225614",
      "postDate": "04/18/2023 10:00:00",
      "content": "<p>For sure you need to use slice, for example my best submission for score 0.63 uses only 7 slices</p>",
      "rawMarkdown": "For sure you need to use slice, for example my best submission for score 0.63 uses only 7 slices",
      "votes": null
    },
    {
      "id": "2226850",
      "postDate": "04/19/2023 09:27:28",
      "content": "<p>which part of slice you took?</p>",
      "rawMarkdown": "which part of slice you took?",
      "votes": null
    },
    {
      "id": "2227015",
      "postDate": "04/19/2023 13:08:48",
      "content": "<p>On the notebook shared with 0.41 score (<a href=\"https://www.kaggle.com/code/tanakar/2-5d-segmentaion-baseline-training)\" target=\"_blank\">https://www.kaggle.com/code/tanakar/2-5d-segmentaion-baseline-training)</a>, the author used 6 slices : index 29 to 35</p>",
      "rawMarkdown": "On the notebook shared with 0.41 score (https://www.kaggle.com/code/tanakar/2-5d-segmentaion-baseline-training), the author used 6 slices : index 29 to 35",
      "votes": null
    },
    {
      "id": "2227247",
      "postDate": "04/19/2023 15:34:24",
      "content": "<p>That's interesting. Do you know what is the heuristic for choosing the 6 slices?</p>",
      "rawMarkdown": "That's interesting. Do you know what is the heuristic for choosing the 6 slices?",
      "votes": null
    },
    {
      "id": "2227622",
      "postDate": "04/19/2023 21:58:04",
      "content": "<p>No I don't know, maybe it's just the result of an experiment because in the calculation, the author starts from half of 65 with a range of more or less n=3.<br>\nI don't know if it would work to try to train a small model on each slice individually to see which are the most informative (best scores)…</p>",
      "rawMarkdown": "No I don't know, maybe it's just the result of an experiment because in the calculation, the author starts from half of 65 with a range of more or less n=3.\nI don't know if it would work to try to train a small model on each slice individually to see which are the most informative (best scores)...",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2225614,
      "author_name": "maksimovka",
      "author_url": "",
      "post_date": "04/18/2023 10:00:00",
      "content": "<p>For sure you need to use slice, for example my best submission for score 0.63 uses only 7 slices</p>",
      "votes": null,
      "replies": [
        {
          "id": 2226850,
          "author_name": "lucario129",
          "author_url": "",
          "post_date": "04/19/2023 09:27:28",
          "content": "<p>which part of slice you took?</p>",
          "votes": null,
          "replies": [
            {
              "id": 2227015,
              "author_name": "davidrouyre",
              "author_url": "",
              "post_date": "04/19/2023 13:08:48",
              "content": "<p>On the notebook shared with 0.41 score (<a href=\"https://www.kaggle.com/code/tanakar/2-5d-segmentaion-baseline-training)\" target=\"_blank\">https://www.kaggle.com/code/tanakar/2-5d-segmentaion-baseline-training)</a>, the author used 6 slices : index 29 to 35</p>",
              "votes": null,
              "replies": [
                {
                  "id": 2227247,
                  "author_name": "yassinealouini",
                  "author_url": "",
                  "post_date": "04/19/2023 15:34:24",
                  "content": "<p>That's interesting. Do you know what is the heuristic for choosing the 6 slices?</p>",
                  "votes": null,
                  "replies": [
                    {
                      "id": 2227622,
                      "author_name": "davidrouyre",
                      "author_url": "",
                      "post_date": "04/19/2023 21:58:04",
                      "content": "<p>No I don't know, maybe it's just the result of an experiment because in the calculation, the author starts from half of 65 with a range of more or less n=3.<br>\nI don't know if it would work to try to train a small model on each slice individually to see which are the most informative (best scores)…</p>",
                      "votes": null,
                      "replies": []
                    }
                  ]
                }
              ]
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2225608": "Since I have limited resources, which include only Kaggle and Colab, concatenating 65 images has become a challenge due to the exceeding RAM memory. To overcome this, I am currently reducing the size of each image by a third and converting them into uint8 format. or should I take a segment of slices from in between",
    "2225614": "For sure you need to use slice, for example my best submission for score 0.63 uses only 7 slices",
    "2226850": "which part of slice you took?",
    "2227015": "On the notebook shared with 0.41 score (https://www.kaggle.com/code/tanakar/2-5d-segmentaion-baseline-training), the author used 6 slices : index 29 to 35",
    "2227247": "That's interesting. Do you know what is the heuristic for choosing the 6 slices?",
    "2227622": "No I don't know, maybe it's just the result of an experiment because in the calculation, the author starts from half of 65 with a range of more or less n=3.\nI don't know if it would work to try to train a small model on each slice individually to see which are the most informative (best scores)..."
  },
  "source": "meta"
}