{
  "id": 415584,
  "title": "Need Help on YoloV8 inference",
  "url": "/competitions/hubmap-hacking-the-human-vasculature/discussion/415584",
  "author_name": "",
  "post_date": "2023-06-07T06:00:48.865032Z",
  "votes": 1,
  "comment_count": 10,
  "views": 0,
  "content": "<p>I need help in making a working submission with my yolov8. All my inference using yolov8 has been failing to score while being successfully run before submission.</p>\n<p><a href=\"https://www.kaggle.com/code/alabibojesomo/inference-yolo/edit/run/131717866\" target=\"_blank\">https://www.kaggle.com/code/alabibojesomo/inference-yolo/edit/run/131717866</a></p>",
  "messages": [
    {
      "id": "2290812",
      "postDate": "06/07/2023 06:00:48",
      "content": "<p>I need help in making a working submission with my yolov8. All my inference using yolov8 has been failing to score while being successfully run before submission.</p>\n<p><a href=\"https://www.kaggle.com/code/alabibojesomo/inference-yolo/edit/run/131717866\" target=\"_blank\">https://www.kaggle.com/code/alabibojesomo/inference-yolo/edit/run/131717866</a></p>",
      "rawMarkdown": "I need help in making a working submission with my yolov8. All my inference using yolov8 has been failing to score while being successfully run before submission.\n\nhttps://www.kaggle.com/code/alabibojesomo/inference-yolo/edit/run/131717866",
      "votes": null
    },
    {
      "id": "2293378",
      "postDate": "06/09/2023 06:20:40",
      "content": "<p>When I used YOLOv8n-seg(smallest one), my submission completed successfully.<br>\nOn the other hands, submission failed when other models were used</p>",
      "rawMarkdown": "When I used YOLOv8n-seg(smallest one), my submission completed successfully.\nOn the other hands, submission failed when other models were used",
      "votes": null
    },
    {
      "id": "2293727",
      "postDate": "06/09/2023 12:38:07",
      "content": "<p>Are you willing to share you inference nodebook？ I also met the inference problems when using Yolov8.</p>",
      "rawMarkdown": "Are you willing to share you inference nodebook？ I also met the inference problems when using Yolov8.",
      "votes": null
    },
    {
      "id": "2293972",
      "postDate": "06/09/2023 16:05:47",
      "content": "<p>you should make your notebook public, so that others could check and comment.</p>",
      "rawMarkdown": "you should make your notebook public, so that others could check and comment.",
      "votes": null
    },
    {
      "id": "2294263",
      "postDate": "06/09/2023 23:20:04",
      "content": "<p>I just made the notebook public. looking forward to any help i can get</p>",
      "rawMarkdown": "I just made the notebook public. looking forward to any help i can get",
      "votes": null
    },
    {
      "id": "2297803",
      "postDate": "06/12/2023 20:18:37",
      "content": "<p>Anyone resolve this? I specifically have an issue with no masks being generated during inference when object (blood_vessel) is detected in the one test image we have. Any input would be appreciated!</p>",
      "rawMarkdown": "Anyone resolve this? I specifically have an issue with no masks being generated during inference when object (blood_vessel) is detected in the one test image we have. Any input would be appreciated!",
      "votes": null
    },
    {
      "id": "2298825",
      "postDate": "06/12/2023 20:50:32",
      "content": "<p>Mhh, your submission file looks fine. <br>\nBut have you checked that it's the case when predicting on more than one image? <br>\nI.e. try and duplicate the test image (or predict on the training images). </p>",
      "rawMarkdown": "Mhh, your submission file looks fine. \nBut have you checked that it's the case when predicting on more than one image? \nI.e. try and duplicate the test image (or predict on the training images).",
      "votes": null
    },
    {
      "id": "2300403",
      "postDate": "06/13/2023 07:25:55",
      "content": "<p>I think I've found a possible issue. Your model can fail to make any detections, in which case your prediction.masks object will be None. It leads to errors, so you should add a check for it.</p>\n<p>I've had the same problem. It works now, with the check.</p>",
      "rawMarkdown": "I think I've found a possible issue. Your model can fail to make any detections, in which case your prediction.masks object will be None. It leads to errors, so you should add a check for it.\n\nI've had the same problem. It works now, with the check.",
      "votes": null
    },
    {
      "id": "2315216",
      "postDate": "06/23/2023 23:22:00",
      "content": "<p>i am pretty sure you gave up with v8, but that here is your problem code:</p>\n<ul>\n<li>use a generator instead of appending to list, it save so much vram</li>\n<li>handle empty pred</li>\n<li>verify weird coordinate when creating a polygon (&gt;=3 point) &lt;-- this</li>\n</ul>",
      "rawMarkdown": "i am pretty sure you gave up with v8, but that here is your problem code:\n- use a generator instead of appending to list, it save so much vram\n- handle empty pred\n- verify weird coordinate when creating a polygon (>=3 point) <-- this",
      "votes": null
    },
    {
      "id": "2315909",
      "postDate": "06/24/2023 14:10:55",
      "content": "<p>Coul you elaborate more on: \"verify weird coordinate when creating a polygon (&gt;=3 point) &lt;-- this\"?</p>",
      "rawMarkdown": "Coul you elaborate more on: \"verify weird coordinate when creating a polygon (>=3 point) <-- this\"?",
      "votes": null
    },
    {
      "id": "2316450",
      "postDate": "06/25/2023 00:17:26",
      "content": "<p>sorry i am not educated so i am maybe wrong: you need more than 3 point not on the same line to create a polygon <br>\nso me i just check if length coordinate of segmentation &gt;= 3 in the get result function<br>\nit seems logical to me but i modified a lot this inference notebook so maybe its broken (but i least it can be submitted)</p>",
      "rawMarkdown": "sorry i am not educated so i am maybe wrong: you need more than 3 point not on the same line to create a polygon \nso me i just check if length coordinate of segmentation >= 3 in the get result function\nit seems logical to me but i modified a lot this inference notebook so maybe its broken (but i least it can be submitted)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2293378,
      "author_name": "shinyatakaramoto",
      "author_url": "",
      "post_date": "06/09/2023 06:20:40",
      "content": "<p>When I used YOLOv8n-seg(smallest one), my submission completed successfully.<br>\nOn the other hands, submission failed when other models were used</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2293727,
      "author_name": "kingjohnson",
      "author_url": "",
      "post_date": "06/09/2023 12:38:07",
      "content": "<p>Are you willing to share you inference nodebook？ I also met the inference problems when using Yolov8.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2293972,
      "author_name": "dragonzhang",
      "author_url": "",
      "post_date": "06/09/2023 16:05:47",
      "content": "<p>you should make your notebook public, so that others could check and comment.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2294263,
          "author_name": "alabibojesomo",
          "author_url": "",
          "post_date": "06/09/2023 23:20:04",
          "content": "<p>I just made the notebook public. looking forward to any help i can get</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 2297803,
      "author_name": "kevinwonjunecho",
      "author_url": "",
      "post_date": "06/12/2023 20:18:37",
      "content": "<p>Anyone resolve this? I specifically have an issue with no masks being generated during inference when object (blood_vessel) is detected in the one test image we have. Any input would be appreciated!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2298825,
      "author_name": "fnands",
      "author_url": "",
      "post_date": "06/12/2023 20:50:32",
      "content": "<p>Mhh, your submission file looks fine. <br>\nBut have you checked that it's the case when predicting on more than one image? <br>\nI.e. try and duplicate the test image (or predict on the training images). </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2300403,
      "author_name": "vyacheslavberezhnoy",
      "author_url": "",
      "post_date": "06/13/2023 07:25:55",
      "content": "<p>I think I've found a possible issue. Your model can fail to make any detections, in which case your prediction.masks object will be None. It leads to errors, so you should add a check for it.</p>\n<p>I've had the same problem. It works now, with the check.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2315216,
      "author_name": "iraqbot",
      "author_url": "",
      "post_date": "06/23/2023 23:22:00",
      "content": "<p>i am pretty sure you gave up with v8, but that here is your problem code:</p>\n<ul>\n<li>use a generator instead of appending to list, it save so much vram</li>\n<li>handle empty pred</li>\n<li>verify weird coordinate when creating a polygon (&gt;=3 point) &lt;-- this</li>\n</ul>",
      "votes": null,
      "replies": [
        {
          "id": 2315909,
          "author_name": "rapela",
          "author_url": "",
          "post_date": "06/24/2023 14:10:55",
          "content": "<p>Coul you elaborate more on: \"verify weird coordinate when creating a polygon (&gt;=3 point) &lt;-- this\"?</p>",
          "votes": null,
          "replies": [
            {
              "id": 2316450,
              "author_name": "iraqbot",
              "author_url": "",
              "post_date": "06/25/2023 00:17:26",
              "content": "<p>sorry i am not educated so i am maybe wrong: you need more than 3 point not on the same line to create a polygon <br>\nso me i just check if length coordinate of segmentation &gt;= 3 in the get result function<br>\nit seems logical to me but i modified a lot this inference notebook so maybe its broken (but i least it can be submitted)</p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2290812": "I need help in making a working submission with my yolov8. All my inference using yolov8 has been failing to score while being successfully run before submission.\n\nhttps://www.kaggle.com/code/alabibojesomo/inference-yolo/edit/run/131717866",
    "2293378": "When I used YOLOv8n-seg(smallest one), my submission completed successfully.\nOn the other hands, submission failed when other models were used",
    "2293727": "Are you willing to share you inference nodebook？ I also met the inference problems when using Yolov8.",
    "2293972": "you should make your notebook public, so that others could check and comment.",
    "2294263": "I just made the notebook public. looking forward to any help i can get",
    "2297803": "Anyone resolve this? I specifically have an issue with no masks being generated during inference when object (blood_vessel) is detected in the one test image we have. Any input would be appreciated!",
    "2298825": "Mhh, your submission file looks fine. \nBut have you checked that it's the case when predicting on more than one image? \nI.e. try and duplicate the test image (or predict on the training images).",
    "2300403": "I think I've found a possible issue. Your model can fail to make any detections, in which case your prediction.masks object will be None. It leads to errors, so you should add a check for it.\n\nI've had the same problem. It works now, with the check.",
    "2315216": "i am pretty sure you gave up with v8, but that here is your problem code:\n- use a generator instead of appending to list, it save so much vram\n- handle empty pred\n- verify weird coordinate when creating a polygon (>=3 point) <-- this",
    "2315909": "Coul you elaborate more on: \"verify weird coordinate when creating a polygon (>=3 point) <-- this\"?",
    "2316450": "sorry i am not educated so i am maybe wrong: you need more than 3 point not on the same line to create a polygon \nso me i just check if length coordinate of segmentation >= 3 in the get result function\nit seems logical to me but i modified a lot this inference notebook so maybe its broken (but i least it can be submitted)"
  },
  "source": "meta"
}