{
  "id": 614726,
  "title": "img_left and img_right?",
  "url": "/competitions/csiro-biomass/discussion/614726",
  "author_name": "",
  "post_date": "2025-11-06T00:46:34.810002400Z",
  "votes": 1,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Why do some (or most of) notebooks process input as <code>[img_left, img_right]</code> format, as shown in the code below. What is the basis for this?</p>\n<pre><code>height, width, _ = image.shape\nmid_point = width // \nimg_left = image[:, :mid_point]\nimg_right = image[:, mid_point:]\n</code></pre>",
  "messages": [
    {
      "id": "3311874",
      "postDate": "11/06/2025 00:46:34",
      "content": "<p>Why do some (or most of) notebooks process input as <code>[img_left, img_right]</code> format, as shown in the code below. What is the basis for this?</p>\n<pre><code>height, width, _ = image.shape\nmid_point = width // \nimg_left = image[:, :mid_point]\nimg_right = image[:, mid_point:]\n</code></pre>",
      "rawMarkdown": "Why do some (or most of) notebooks process input as `[img_left, img_right]` format, as shown in the code below. What is the basis for this?\n```python\nheight, width, _ = image.shape\nmid_point = width // 2\nimg_left = image[:, :mid_point]\nimg_right = image[:, mid_point:]\n```",
      "votes": null
    },
    {
      "id": "3311877",
      "postDate": "11/06/2025 00:49:43",
      "content": "<p>I too do not understand this, infact using entire image gives better results.</p>",
      "rawMarkdown": "I too do not understand this, infact using entire image gives better results.",
      "votes": null
    },
    {
      "id": "3311900",
      "postDate": "11/06/2025 02:11:48",
      "content": "<p>To make it become two square pic and compress it without deformation </p>",
      "rawMarkdown": "To make it become two square pic and compress it without deformation",
      "votes": null
    },
    {
      "id": "3311904",
      "postDate": "11/06/2025 02:21:00",
      "content": "<p>Thanks, just as I thought.😁</p>",
      "rawMarkdown": "Thanks, just as I thought.😁",
      "votes": null
    },
    {
      "id": "3311905",
      "postDate": "11/06/2025 02:21:46",
      "content": "<p>Thanks, I'll give it a try.</p>",
      "rawMarkdown": "Thanks, I'll give it a try.",
      "votes": null
    },
    {
      "id": "3318101",
      "postDate": "11/11/2025 02:32:02",
      "content": "<p>I think it's because most of the pretrained models have input shapes like 224x224, 512x512, etc and not 1000x2000 like this contest. Interested to see if splitting, inputting whole, or some other technique comes out on top though.</p>",
      "rawMarkdown": "I think it's because most of the pretrained models have input shapes like 224x224, 512x512, etc and not 1000x2000 like this contest. Interested to see if splitting, inputting whole, or some other technique comes out on top though.",
      "votes": null
    },
    {
      "id": "3322109",
      "postDate": "11/13/2025 10:27:14",
      "content": "<p>thanks,helpful</p>",
      "rawMarkdown": "thanks,helpful",
      "votes": null
    },
    {
      "id": "3388151",
      "postDate": "01/08/2026 10:10:05",
      "content": "<p>Sometimes, this approach is better, and sometimes the entire picture is better. It depends on what model you use</p>",
      "rawMarkdown": "Sometimes, this approach is better, and sometimes the entire picture is better. It depends on what model you use",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3311877,
      "author_name": "paritoshtripathi5",
      "author_url": "",
      "post_date": "11/06/2025 00:49:43",
      "content": "<p>I too do not understand this, infact using entire image gives better results.</p>",
      "votes": null,
      "replies": [
        {
          "id": 3311905,
          "author_name": "wqx20000115",
          "author_url": "",
          "post_date": "11/06/2025 02:21:46",
          "content": "<p>Thanks, I'll give it a try.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 3311900,
      "author_name": "wliilamsam",
      "author_url": "",
      "post_date": "11/06/2025 02:11:48",
      "content": "<p>To make it become two square pic and compress it without deformation </p>",
      "votes": null,
      "replies": [
        {
          "id": 3311904,
          "author_name": "wqx20000115",
          "author_url": "",
          "post_date": "11/06/2025 02:21:00",
          "content": "<p>Thanks, just as I thought.😁</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 3322109,
          "author_name": "handudu",
          "author_url": "",
          "post_date": "11/13/2025 10:27:14",
          "content": "<p>thanks,helpful</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 3318101,
      "author_name": "msthil",
      "author_url": "",
      "post_date": "11/11/2025 02:32:02",
      "content": "<p>I think it's because most of the pretrained models have input shapes like 224x224, 512x512, etc and not 1000x2000 like this contest. Interested to see if splitting, inputting whole, or some other technique comes out on top though.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3388151,
      "author_name": "sixueyang",
      "author_url": "",
      "post_date": "01/08/2026 10:10:05",
      "content": "<p>Sometimes, this approach is better, and sometimes the entire picture is better. It depends on what model you use</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3311874": "Why do some (or most of) notebooks process input as `[img_left, img_right]` format, as shown in the code below. What is the basis for this?\n```python\nheight, width, _ = image.shape\nmid_point = width // 2\nimg_left = image[:, :mid_point]\nimg_right = image[:, mid_point:]\n```",
    "3311877": "I too do not understand this, infact using entire image gives better results.",
    "3311900": "To make it become two square pic and compress it without deformation",
    "3311904": "Thanks, just as I thought.😁",
    "3311905": "Thanks, I'll give it a try.",
    "3318101": "I think it's because most of the pretrained models have input shapes like 224x224, 512x512, etc and not 1000x2000 like this contest. Interested to see if splitting, inputting whole, or some other technique comes out on top though.",
    "3322109": "thanks,helpful",
    "3388151": "Sometimes, this approach is better, and sometimes the entire picture is better. It depends on what model you use"
  },
  "source": "meta"
}